Good afternoon, everyone. Welcome to our second annual ShotSpotter Investor and Analyst Day. We couldn't be more thrilled to have as many of you come out and be online to give us the opportunity to share the progress we've made in the business over the past year. My name is Ralph Clark. I'm ShotSpotter's President and CEO. I want to personally thank, with a great deal of gratitude, the many of you that have been so supportive of me and the company since we've gone public two and a half years ago. I think your feedback has been incredibly important to me and the company as we conducted ourselves as a public company, hopefully somewhat successfully. We're really excited about what we're going to be covering with you today. A couple of favors. Let me get my pointer here. Great.
We have a fairly full day, and I've got two favors to ask of you. One favor you're going to be very familiar with. This has to do with our forward-looking statements clause here. We're going to be together for about three and a half hours, and I'm very sure that during the course of the presentations, you're going to hear some forward-looking statements, and those come, of course, with the appropriate clauses. We want to be as open and transparent as we can be within the bounds of good taste as a public company. We would encourage you, to the extent that you are going to hear some forward-looking statements, we encourage you to go back and review our formerly SEC filed 10-K and our most recent 10-Q and review that in detail.
Pay particular close attention to risk factors. We want you to balance your notion of the business from the opportunities you're going to hear about today, along with the risk factors of the business, so that you can be very well-informed as a potential investor. The second favor I want to ask of you is, given the fact that we have so many people that are online, and we have a lot of material to cover, we generally like to have these be conversations with a little bit of back and forth. Of course, that's very difficult for folks to kind of follow through on the pace of the slides as we do that. We're just going to ask you to hold your questions to the end. We're going to make very sure that we have lots of time. Do I need this?
Yeah.
Okay, great. Okay, thank you. We're going to have lots of time to be able to answer your questions towards the end. Does that sound reasonable to everybody? Okay, awesome. Let's go ahead and get started. This is the agenda that we want to cover. You're going to hear from the entire senior leadership team. I got to tell you, I couldn't be more proud of the team that we've built here at ShotSpotter. We have a very experienced senior leadership team that's very skilled and very passionate about the work that we do. You're going to hear from every single functional organization, and I think you're going to walk away fairly impressed.
I know you see Alan and I a lot on the road, you're going to have the opportunity to hear from our cohorts on the senior leadership team, you're going to be impressed with their functional leadership. Hopefully, you're going to also be equally impressed with the way we are very cohesive and of one mind and approach in terms of how we move through the market. We're going to do a bit of a deep dive on the technology piece. We've allocated about 40 minutes or so for both Paul Ames and Rob Calhoun, who is one of the founders of the company. He's one of our principal architects.
They're going to do a bit of a deep dive on the technology. The reason for that is we get a lot of questions from you all and other investors about what makes our specialized solutions so unique and proprietary. They're going to take you on a little bit of a journey on math and physics. I just ask you to kind of stay with it, and I think you'll definitely benefit from a better understanding about our technology and how it works and what makes it so special. Then everyone else will kind of do their thing. Again, we'll hold questions to the very end. Then we're going to leave some time to be able to go through a formal tour of the Incident Review Center and a Missions demo. Where do we start in terms of the company?
I always like to start with the question of why or what the purpose of the company is. At ShotSpotter, we feel like we do some very important and impactful work. The thing that it is not, is not the work of individuals or an individual company. Although the original inspiration for the company came from an individual, our founder, Dr. Bob Showen, who had a vision that he could use technology to impact the issue of gun violence. We're building on that inspiration and built a purpose-focused company that bring together not just the company and our partners, but it also brings together customers and investors, all aligned around our sense of purpose. The bigger we can make that purpose focus, the more alignment we can drive.
That purpose is really about aiding law enforcement in becoming more relevant to and important to communities and helping them address the issues of gun violence, simply said. There's more of a formal purpose statement there. At the end of the day, it's really about us being of service to police departments so that they can be in service to communities to prevent and reduce gun violence. Because we have this strong alignment across the ecosystem, it enables us to be aligned. In that alignment comes a lot of power. We're able to punch in well above our weight. Being the relatively small company that we are, we have a tremendous impact because of the way that we're all working together around focusing in on this particular mission. That's critically important to us. Now, this issue of gun violence is a very big issue.
Again, it's the thing that inspired Dr. Bob to take this incredible invention insight that he had to leave his very comfortable job at Stanford Research Institute to develop the ShotSpotter technology. We all read about gun violence quite a bit in the news, and it's a very big problem. On average, there's about 13,000 people that are killed a year due to the criminal use of a gun. Just think about that for a moment. That's quite a sobering statistic to think that 13 people's lives can be taken on an annual basis due to the issue of gun violence. That's an incredibly big number. If that number isn't sobering enough, there's another 75,000+ folks that are wounded or encounter gunshot wounds, or GSWs, as we refer to them internally.
The thing about a GSW is, they're extremely expensive because oftentimes the victims of gunshot wounds require very expensive ongoing medical support. When you include the loss of productivity, the loss of quality of life, the cost of a GSW is significant. Combining homicides with GSWs, and if you're lucky enough to avoid either one of those particular poor outcomes, it's quite likely that you're not going to be able to avoid the emotional trauma that comes along with living in these environments that have ongoing, persistent gun violence. We refer to that as vicarious victimization. That's particularly true if you're a child or adolescent, because we know from scientific research that children that are exposed to ongoing trauma, their brains are rewired. The process of rewiring those brains drives all manner of very negative outcomes, both physiological outcomes and emotional outcomes.
Think about impulse control issues, how people deal with trauma through self-medication, drug use, alcohol abuse. If you have very poor control over impulses, you're more likely to engage in all manner of risky behavior, teenage pregnancies, learning difficulties, the whole nine yards. When you add all those costs up together, it's the thing that compels the market to demand proven solutions to address this very costly problem. Again, it's direct costs, it's indirect costs, it's societal costs, it's economic costs, and it's big. It almost seems intractable, because part of the problem is that 80% of gunfire incidents, and this has been researched by Brookings Institution, 80% of gunfire incidents go unreported. The traditional narrative is guns are fired, no call to 911.
When the calls are made to 911, they're oftentimes made five minutes after the shot has been fired, and they almost always never have a precise location in which to direct law enforcement resources, which means we're sending our most precious resources pretty much out on a wild goose chase. That is the narrative of the analog world of this particular problem. What makes this problem encouraging from our point of view is that although we talk about gun violence on the national policy level, the fact of the matter is gun violence is local. Let me repeat that. Gun violence is local, and not only is gun violence local, gun violence is hyperlocal.
What we're talking about across all the cities that are dealing with gun violence, it's a portion of the city that has the ongoing persistent gun violence, which makes it potentially solvable, because it's not like you have to boil the ocean across the entire United States to solve the gun violence problem, nor do you have to boil the ocean across an entire city to solve the gun violence problem. The gun violence problem is really about hyperlocality within a city that makes it somewhat potentially addressable, and we're going to talk about how we address that problem with the ShotSpotter solution. The other insight that we know, and this is well-researched from a criminology research point of view, is the Pareto effect. How many of you are familiar with Pareto? The economic rule of Pareto.
Pareto says this kind of 1% of something causes 80% of something else. In the case of violent crime, it's very much Pareto. It's a Pareto optimum issue. It's very few serial trigger pullers that drive most of the gun violence in any one of these hyper small targeted locations. That means any potential solution to address this issue, if they're focused around focused deterrence, where you're trying to have a law enforcement intervention with the very few people that drive the vast majority of the problem, you can drive impact. That's a bit encouraging for us. Although we don't ever try to take singular credit for solving the gun violence issue, we can solve the first mile of the gun violence problem, which has to do with the detection piece. Again, let's refer back to that 80%.
The typical narrative is guns are fired, no calls made, no police response. 20% of the time in the analog world, the call is made. It's made late. The call goes something like, "Hi, I think I heard a gunshot." The operator asks, "Where?" I say, "I'm not exactly sure. I think it's out there somewhere." There's all manner of cognitive bias that comes in with where I think I heard the gunshot. It's probably down the street to the neighbor that I don't like very much. That's always where I hear the gunshot, right? It's not very accurate. I just want for a moment for you to kind of close your eyes and think about the possibility of this different world.
What if I told you there was a scenario in this digital world where every time there was a gunshot, there was a digital call for service. The beauty of this digital call is that the digital call happened within 30 to 45 seconds of every time a gun is fired. That digital call was very precise in the location which you said the gunshot happened. It had a lot of metadata associated with it, so not only the location, but a brief description about the nature of the gunfire event. Number of rounds. Is it a multi-shooter? Is it a single shooter? It's an automatic weapon. Is it a semi-automatic weapon? Is the shooter moving, et cetera. Imagine this digital alert being presented in the form of a map with a dot.
Again, the precision is around 20 meters. It happens every single time a shot is fired. Because it's digital, it can be consumed digitally, which means that that digital call for service can not only go to a patrol officer, it can go to a PSAP organization, which is the answering point for 911 calls. It can be ingested by video management systems. It can go a number of different ways, as opposed to the analog manual process that goes directly into a PSAP, into a call for service that is very difficult to repurpose. It's fast, it's precise, it's omnipotent in a way because it's happening not 20%, 10% of the time, it's happening 95% of the time. Although our SLA standard is 90%, I want to make sure I make that point as well. It's happening frequently, it's very precise.
As a result of getting that now, you can start the process of the value chain. We talked earlier about the fact that if you're going to make progress with gun violence, it really has to do around focused deterrence in terms of really focusing your intervention resources on the very few people that are driving the problem. What we do is we solve that first leg of the problem. We go from the 80% no calls for service to 90%+ calls for service that's precise, fast, digital, can be repurposed in many different ways. It's a digitization of an analog manual process and very efficient as a result. That's what we take singular credit for at ShotSpotter. Those detections then lead to responses.
From a best practices point of view, law enforcement now has the confidence to be able to respond very precisely and directly to these events. As a result, if they're not dealing with a perpetrator or aiding a victim, they're much more likely to be able to recover physical forensic evidence in the form of shell casings that can be run through a fingerprinting system for shell casings. They're much more likely to be able to interview witnesses with ground truth about what happened. What this does is it sets up very informed leverage investigations. Now in our analog world that I described before, what people are doing is they're investigating homicides. The fact of the matter is that these serial shooters are leaving all manner of evidence around for us to use if we're able to get to these events quickly with ground truth.
We now have the ability to investigate not just homicides and misinvestigate GSW events. We have the opportunity to investigate shootings, and the more investigation of shootings we can create, the more likely it is we can identify the serial shooter, which then leads to the intervention. Interventions just don't have to be law enforcement interventions. They can be social services interventions as well. One of the more interesting stories I'm really proud of is that one of our customers figured out the culmination of ShotSpotter data really informed them around this hyper location capability or hyper location aspect of gun violence, where they saw that from a spatial and temporal point of view, it was happening in this one very specific part of the neighborhood, Thursday, Friday, and Saturday night during the summer.
What they figured out was they said, "Hey, look, what if we opened up that gym that is literally right in the epicenter of where we happen to have this gun violence problem? Let's put a little bit of money around establishing a midnight basketball league for Thursday, Friday, and Saturday." Okay. That's what they did. What that enabled was the community to reclaim the space to make it a lot more difficult for serial shooters to operate. The beauty of this is they were able to prevent and reduce violence in that specific part of the neighborhood without arresting a single individual and using very cheap resources to do that. What were those resources? They were community resources. Guess what? Community resources, they don't belong to unions. They don't require vacation. They don't require benefits. They're extremely cheap, and they're super effective.
That's a very creative way to think about using data to drive interventions. A lot of our interventions are going to be traditional law enforcement interventions, and you just need that because, for some knuckleheads, the only thing that they're going to understand is when you cuff them. When you're able to get the culmination of investigative leads that we're now able to present, because now you're showing up to 100% of the gunshot activities as opposed to 10%-20% of the activities with you investigating homicide versus shootings. You're more likely to have a successful law enforcement intervention. As you get those interventions, you drive deterrence. You drive deterrence because you take that Pareto off of the street. You reduce the conflict that they have with the other shooter. The word gets around that there's a new game in town.
In fact, one of our customers calls it a game changer from that point of view. You start to see the behavior change and the denormalization of gun violence, which then leads to healthier communities. The healthier communities is the thing that systemically prevents and reduce gun violence. All the law enforcement resources in the world, all the brilliant technology, like the technology that we offer in the world, cannot systemically prevent and reduce gun violence. We do our part, our part leads to the value chain that gets to the healthier communities. It's the healthy community that prevents and reduces gun violence. For our customers that understand this, they really embrace this concept, they see amazing results. Now, the challenge for us is, of course, when we see these results, we can't, as a company, singularly say that ShotSpotter caused all this, right?
Guess what caused it? I mean, we did our part on the front end. Police did their part in the middle, then the community does their part on the end, I guess. The prosecution resources over here do their part. This is, again, why purpose is so important. This is all about a collaborative effort between different technologies, different resources, different organizations, all aligned around the principle of providing public safe spaces for at-risk, underserved communities that, by right, deserve it 100%. That's what we're doing here. That's why our work is so important and so impactful from a societal point of view. Of course, we're here because this is a business too. We very much believe in doing well by doing good. Let's talk a little bit about where we are as a business.
I think many of you know we're deployed in about 100 or so cities, mostly domestically, incorporating about 700 mi. About on average, seven square miles per customer. That involves us deploying about 15,000 sensors. These sensors are very sophisticated sensors that collaborate with one another to produce these vetted alerts. Last year, 2018, we published over 100,000 vetted gunshot alerts across the world. Along the way, we've done a lot of work in terms of the technology and our intellectual property, and I can proudly say we've amassed a very powerful patent portfolio of 34 issued patents. For a company our size of just over 100 employees, to think about the fact that we have one patent for every three people is really pretty impressive.
I wouldn't count as one of the three people because I'm not that smart, but you're going to hear about some of the people that have done the work around that. It's an incredible technology that's producing these business results. 38% CAGR growth from 2015 to 2019 estimated. Part of the thing that's driving those hard assets to the left of the screen is the soft assets around developing a place that's purpose-driven, where employees feel good about the work, right? We're very keen on developing culture. Culture is our operating system that allows us to have a bigger impact than what almost any other 100-person company could drive. It's that culture right here. Okay? As aspirational as we are, and we need to be aspirational because it motivates people, it brings people together.
We also have to have line-of-sight objectives as well, because we do have a business to run. Our line-of-sight objectives on the business side is to grow this business to be a $100 million revenue business. That revenue, we want it to be recurring, diversified and profitable. The path by which we make that happen is, the first thing we have to do is protect and maintain what we think will be approximately $43 million of annual recurring revenue, beginning 2020 or end of 2019. You're going to hear a lot from Nasim and Paul about what we do on the customer support IRC and interestingly, on our customer success organizations, all optimized around how we protect that $43 million. The next thing we want to do is add about 600 sq mi domestically.
100 of those 600 mi, we expect to come from our existing customer relationships. That 100 or so customers that has covering about 700 sq mi, those 100 customers will represent 800 sq mi if we're successful in adding approximately 100 sq mi in expansion opportunities. We also need to get another 500 mi from new customers, representing about 100 or so cities. In total, 100 plus 500 equals 600. You can multiply that times our annual subscription price to kind of get to the additional annual recurring revenue that's on top of the $43 million. We're very excited about the opportunity to grow our international business. International for us basically represents the Caribbean, Latin America and South Africa. We believe we have what it takes in terms of the demand pull.
You're going to hear from Jon later on about some of the macroeconomic drivers in Latin America, primarily, that's driving the demand for compelling solutions like ShotSpotter. That gives us a lot of confidence that we're going to be resourced to get this to, at the $100 million run rate, to be about 15% of our business. On the Missions and Security side combined, we'd like to grow that business to about $10 million in revenue. Missions is our inorganic acquisition that extended our platform beyond traditional gunshot detection. Security, of course, uses our traditional gunshot detection technology, but it's a different buying center focused on more higher ed campuses and corporate campuses. Different buying center, slightly different use case, more around occasional gunfire or active shooter gunfire, as opposed to ongoing, persistent gun violence that typical city police departments deal with.
We think between missions and security, that represents $10 million. I think if you add all that up, we're knocking on $100 million. That's our medium-term goal as a business to get to. Very achievable, from my point of view. Underneath the covers, what we're doing, I won't spend a lot of time on this. I can give the mic over. I think that's my time. Growth levers underneath it. We're going to continue to invest in technological innovation and seek platform extension opportunities as appropriate. We're big believers in NPS, net promoter score and customer onboarding and success, to drive broader, deeper adoption. Because that sets up a virtuous cycle of network effects around reference selling. You're going to hear later about our expanded go-to market capabilities, involving external partnerships as well.
Then we're always going to remain passionate and purposeful about our collective work in making a difference, because again, that's the operating system that fuels this company. With that, let me turn it over to you, Sam. This has got to be coordinated.
Yeah. Great.
Okay, good afternoon. My name is Sam Klepper. I'm finishing my second year running the marketing and product strategy team. It's been an incredible experience for me, very rewarding. The mission of the company is very energizing, the team that we get to work with, just the opportunity to make an impact on society. I'm going to walk you through some of the markets that we serve, and how we're approaching them from a marketing lead generation and product standpoint. I'll start with the total available market, over $1 billion. The core piece of this, of course, is our U.S. domestic Flex products, gunshot detection. We look at this as of the 15,000 law enforcement agencies, 1,400 are dealing with an above average number of homicides driven by gun violence, according to the FBI database.
We're looking at 1,400 domestic cities, given the average coverage area of our current customer base and what we charge, it's about $400,000 per city for about a half a billion dollar piece of the overall addressable market that we have. There's an increasing sense of urgency by many of these cities as gun homicides have been increasing significantly over the last several years. We are able to take that same technology with some very slight modifications and bring that to the international market that Ralph mentioned. This is just looking really at Latin America and South Africa, about 200 cities. These communities within these countries are suffering from significantly higher gun violence than in the United States. It's a more urgent problem there.
Based on the coverage area and what we charge internationally based on some higher costs we have around local data residency and support, it's about a $1 million opportunity per city, that's about $200 million in total addressable market. In terms of the secure campus opportunity, there are about 5,000 college campuses. Again, it's the same technology that we bring to this market. 5,000 college campuses, annual recurring revenue, about $50,000 per campus, which is about $250 million annually. The newest opportunity with ShotSpotter Missions. We think this is a conservative estimate on the total available market. We acquired this product approximately a year ago. We're still learning what the opportunity, how big it really can be. The nice thing about this is that this can address cities that don't even have gun violence problems. They can have property crime problems or other types of problems.
It doesn't have to be limited to gun violence. There are about 1,500 cities we're looking at initially, about 50,000 per city for just under $100 million of the total opportunity there. That adds up to over $1 billion. I will point out that the majority of the market that you see up here, there is really no direct competitive solution that we see in the marketplace that poses a challenge, which is pretty different from probably most of the TAMs that you look at, where there is direct competition and lots of it. How are we attacking this opportunity across these different segments? The most important goal that I had this year with the marketing team was around developing a lead generation engine to bring in new prospects, to revisit old prospects that had gone cold and turn them into warm leads.
We made a lot of progress in 2019. We built the foundation in terms of the technology that we use to automate things and report on how we're doing. We brought in a small team to run this process, to drive campaigns, we brought in some folks who are actually qualifying these prospects. It was an experiment for the first year, but we have some really nice results. What we're seeing are things like in the second half of the year, we found with our SDRs, sales development representatives, these are the folks who call the prospects. They're on the marketing team, saving the sales team time not to deal with prospecting, but instead to just take warm leads and try to close them.
We found that we doubled our number of booked meetings for the sales team in the second half versus the first half. The number of booked meetings per week doubled in the second half of the year. What's very important in this process is that we're getting good leads, qualified leads. We have criteria that we agreed on what a good lead is. We have a 95% qualification rate that we have so far, and most of those leads turn into qualified opportunities going to Salesforce and are part of the pipeline for the company. Now, the engine, we're doing email campaigns, events. We do over 25 events a year. We have inbound interest coming in. It's kind of interesting. There was a Live PD television show. We get leads from that. We're mentioned in Castle, different television shows, and we actually get leads from that.
You got your inbound coming in. All of these are different channels that we get. One thing we haven't done that we'll be doing next year is doing more outbound calling and offline advertising and online advertising, and testing that to see if that can bring in some good leads for us. Obviously, what happens is we try all these different channels. The leads come in, they go through this qualification process with the SDRs. There's a nurturing process because not everyone's ready to buy right away, and then eventually they become opportunities and part of the pipeline. What we're focused on in 2020 is taking kind of phase I we created in 2019 and turning it into phase II. We're going to be a little more sophisticated about the marketing by segment, and I'll go into some details on this.
Based on the city size, for example, there's different ways we can approach it from a sales and marketing standpoint. There's lots of influences on the chief of police. We are understanding better who are those influencers and what are their motivations, and how can we get to them and get our message in front of them that is most important for their concerns, their priorities. We're developing more content. That's been a bit of a limitation on our end. We have a baseline of content. I was able to find someone who has a lot of experience in this area. She actually had higher offers from other companies, but she came here because of the mission, and she'll be starting in January to help drive more compelling content, which is the key to winning over these prospects.
With that content, we'll be able to increase the cadence of our campaigns. Instead of one-offs here and there'll be a whole program of campaigns throughout the year. We'll be starting a lead nurturing program to warm customers up automatically through an automated system, before they get turned over to sales and expanding the actual SDR program with more people. Now I'm just going to dive into a few of these areas a little more deeply. On the market segmentation side, just kind of the high level based on size of the city, and the number of sworn officers in the cities. There's tier 1 through 4 that we focus on. We're looking at really focusing, at a strategic level, customizing our approach to the few tier 1 opportunities that are out there for us to convert.
We have basically VP level and above working with the regional sales directors. There's political aspects that are involved in these cities, to get some additional momentum going for these larger cities. Tier 2 and 3 are kind of bucket together, those are using this core lead gen engine to create the leads, working with the sales directors, for these medium-sized cities. Tier 4, we're looking at having inside sales of lower cost resources come in and focus on those lower tier cities who have different concerns than some of the larger cities, different needs. Just for example, one of them is on onboarding. There are fewer resources in the department to take on the product, and so that will be an emphasis in our messaging that we can support them better through our onboarding process. I mentioned personas and influencers.
These are some of the key influencers that we're aware of, that influence the Chief of Police. We made some progress with hospitals we've talked about in the past. We've made some progress on the community, doing some programs in those couple areas. Areas in 2020 we want to look at are, can the District Attorney be an influencer? To what degree? What are the messages that go to the District Attorney and his or her office? The other area that will be interesting to look at are associations like the Police Executive Research Forum, the Major Chiefs, the Police Foundations. How can we influence them and get our content and get the message out through those very respected associations? I just wanted to give you a couple examples of elements of our campaigns that are coming up.
Las Vegas Metropolitan Police Department, they've had a lot of success with ShotSpotter. They had a nine-month pilot, and they announced in October that they were expanding significantly. They did their own press conference. We were all blown away at the company, at all the data that they used to justify this expansion. We're turning that into a campaign. We simply took the video and with their permission, added a few things to the video to summarize key points, put some overlays on that video. We created a companion article for someone who wanted to read the key points rather than listen to the video itself. We're going to be using this to demonstrate the effectiveness of ShotSpotter. This is a great example of a nurturing campaign.
Someone who's shown some interest, now here's a happy customer who's got data to prove that ShotSpotter is working for them. An additional program we're doing is a more targeted program at prospects who have NIBIN or a CGIC, Crime Gun Intelligence Center. These folks have a machine where they can take the shell casings and get matches to the gun that had fired that bullet, and connect it to an individual, ultimately two witnesses. We know that there is a maximization of that NIBIN machine if you have ShotSpotter, because ShotSpotter's been proven by the Urban Institute to increase the amount of evidence that is found. Depending on the type of crime, it can be two or three times more shell casings are found when used in conjunction with ShotSpotter.
We're going to target the head of this Crime Gun Intelligence Center and the chief of police of cities that don't have ShotSpotter, but they do have NIBIN. We're going to have a webinar with customers who previously did not have ShotSpotter and had NIBIN, and then they took on ShotSpotter, to talk about the dramatic improvement in the results of their NIBIN program, and we'll have some companion articles that go along with that. I want to show you, this is a video for folks who are online. Unfortunately, you won't be able to see it, but you should be able to hear it. We were at IACP, which is the largest chief of police conference in the country, in October, and we brought our video cameras, and we filmed some chiefs who wanted to say some great things about ShotSpotter.
This is just a rough cut of that, but you'll get an idea. In this business, I've been in a lot of different businesses, word-of-mouth is always important. In this business, it's even more important. It's important who says it. Does the marketing guy say it? It doesn't matter. If a chief of police is talking to another current customer of ShotSpotter, chief of police, can talk to a prospective customer, chief of police, that's gold, and tell them about ShotSpotter and what it means to them. The way to scale this, we think, is through video and get the chiefs who are promoters of the product to tell what their thoughts are on it. This is a couple-minute video on that.
[Presentation]
This is an example of content that can really make a difference with chiefs, and this is one view of it. This is in a medley, right? They're all kind of strung together. We have some great interviews that go deeper on particular subjects, so we can slice and dice these videos up and send them out at different parts of the phase of a customer's life cycle. Other things that we're working on for 2020 is just more We've been in business for 20 years. The product works incredibly well. There's noise in the market about maybe it doesn't work well. Just using data. Here's an example of a data sheet we put together recently that gives hard numbers about results agencies have had with the benefit of ShotSpotter being part of their overall solution that they're using to reduce gun violence.
Engaging third parties, independent research, not just us saying it, not just the agency saying it, but an independent research organization. Here's an example, the Urban Institute. They're on the East Coast in D.C., a nonprofit. They conducted, from 2016 to 2019, a study across three cities, you can see Denver, Milwaukee and Richmond, California, on effectiveness. The first thing they did was they published a best practices guide. The conclusion being this can be a very effective technology if used properly, if the police put it on their phones, if the police respond, if the police get out of their car and collect evidence and put it into the NIBIN system, for example. The part of this report talked about it confirming our ability with quantitative numbers around identifying gunshots that were not called into 911. 80% of gunshot incidents are not reported, confirming that.
Improved response times by the police due to the quick nature of the digital 911 system and what I mentioned earlier about the increase in evidence that is recovered when ShotSpotter is employed. I think this is a big part of 2020 and having more academic-oriented proof of results. Here's one issue that I think we've done a decent job getting ahead of. There are several cities who are adopting very stringent surveillance laws. Oakland, San Francisco, locally here are examples. We saw this trend happening, and we thought our system was designed in such a way, very purposefully, to avoid the risk of voice surveillance. It's not designed to do that. We decided to have an independent party look into it and have an audit. They did an audit. They had access to all of our systems, our people, full editorial control of the report.
This came out at the end of July with the key conclusion that there's a very low risk of voice surveillance. We've used this effectively in Oakland, right at the end of the month, last month. We passed the Oakland Privacy Commission process, and Oakland is considered to have the strongest surveillance policy in the U.S., according to the ACLU. I think we're in good shape. Once we can educate a community, a department, about what we actually do to reduce the risk of potential voice surveillance, usually things calm down, and we can move forward with the sale. I'm going to spend a few minutes on ShotSpotter Missions, formerly HunchLab, the acquisition from last year.
The way we approached this is we didn't want to just slap our brand onto the product and just start selling it the way it had been sold previously, the exact same product. We wanted to really gain some key customer insights from agencies to take this from being a very good product to a world-class product. The only way to do that is to have law enforcement, our trusted partners, our existing customers, give us those insights by actually using it. We targeted our current Flex customers versus just going out to everyone in the world and trying to push this product. We intentionally delayed a formal launch.
We had some informal conversations that were discovery conversations to understand how they do patrols, how they deploy their resources, what processes they use, what kind of reporting they had, how the crime analysts were involved, et cetera. We gathered a lot of good information to come up with how we message this product, how we would pitch it, how we price it. That gave us some time to do some things in the product that we knew were very important already without any feedback from customers, which was to integrate our gunfire data into the product, and then to also get some better reporting. We formally launched with what we call an early adopter program. You've probably heard of those things. Usually, a company will launch a product and, at least in this category, they'll offer it for free. We didn't do that.
We did offer some special pricing, and I'll tell you a few more details about that. For a customer to be part of the early adopter program, they had some contractual obligations to us, which were to, number one, provide feedback on a regular basis about what was working, what was not working, their ideas for how to make the product better, and also to be references for us. If they agreed to that, we gave them special EAP pricing, and then we did make an offer to accelerate the urgency, to create urgency and accelerate them signing in this year because we want these customers signed up. We want them deployed to get the learning process started quickly. We offered them one free mile of Flex for one year in a contiguous area.
The contiguous area piece is simply about it's less expensive for us to deploy versus in a completely separate area because we've got sensors that are already nearby. There's fewer sensors that are required for that. That's a very limited, very targeted offer, and it actually ends at the end of this month. That will not continue into 2020. We've had some good progress signing up customers. We have, as of this morning, five deals that have been signed. There's a customer that's been deployed for a couple of months. There's another one deploying this month. There are two more we expect by the end of the year in these last couple of weeks. We expect three of the seven total signed in this year to have multi-year contracts, three years specifically.
We're getting some good interest, and we look forward to seeing more enhancements to this product and more deployments of this product. In terms of product overall, I'm not going to go through a product roadmap with you, but I just thought I'd give you a flavor of what we're doing by talking about the goals. If you look at the four boxes here, from left to right, you're looking at the highest investment level to the lowest investment level. Of course, the highest is U.S. Flex, the biggest part of our business today. It's really not about a bunch of new bells and whistles. There will always be some new user features, but the real focus continues to be getting better and better at location and classification, basically accuracy of the product.
The second piece here mentioned is our Investigator portal project or product that will be renamed Investigate. It's a new version. It's based on Silverlight. We need to upgrade the product, and we have great customer feedback on how to make it even better in this next version. The second area of focus, Missions, I've talked about already, gaining some insights to enhance and differentiate the product. The third area is international Flex. There are some modifications that need to be made. It's important to have local data residency, so there's some operational pieces there that have to happen. There's some communication pieces that need to be modified for the cellular communications. To ensure we're in privacy compliance with each of those countries' regulations. Lastly, it's really seeding new potential future opportunities, making small bets that can turn into commercial enterprise in the future.
One area that we've talked about for the last year or two is there an opportunity for the ShotSpotter Flex product to have some kind of consumer or business service associated with it? What would that look like? What would be the business model for that? I'm going to speak to that in a few minutes. The other area that we're going to invest a little bit of resource in is with the fish bombing. I'll talk to that in a little bit more detail in a second. Finally, we're constantly producing data internally for things like the National Gunfire Index we do every year. We have data customers, and it's a manual process, so we're going to introduce some tools to improve productivity when we do pull data. Those are the overall product goals.
I wanted to spend a minute on ShotSpotter Labs and talk about that fish bombing piece as well. ShotSpotter Labs, you may remember, was established in April of this year, philanthropic arm of ShotSpotter. We have the anti-rhino poaching example in Kruger National Park in South Africa, and we've overcome a lot of the technical challenges there. There's a hostile environment, no electricity. Our goal for Labs is philanthropic, but at the same time, there are opportunities to take the technology innovations and bring them back into the core business. An example there is, with no electricity, we had to figure out how to deploy using solar, and that same kind of solar approach was used in our deployment locally here for a freeway project. That's a good example of taking that innovation into the commercial business. This year, we've proven the efficacy.
Poaching has gone down in the area that we cover from an average of three poaches per month to zero. It provides the possibility of commercial opportunity because the poachers are still looking for those horns. They're going somewhere else, and they're going to the private reserves, the game reserves. We're now getting inbound interest from those game reserves, so that provides potential commercial opportunity for that part of ShotSpotter Labs. The second piece is the fish blasting. First, just recently, we did our first detection. These are underwater detections of bombs that are thrown into the water so that a fisherman can get a lot of fish all at once. The problem there is it destroys, ultimately, the coral reef.
The coral reef is the food supply for the fish, and the fish is the food supply for the people. This is a food security issue. First, I'm going to play what you probably have heard from us before. A gunshot above ground. This is what our first detected fish bomb were from underwater. It's pretty different. These are hydrophones under the water. There's some interesting pieces around this technology where sound underwater propagates much more quickly and more widely, so you need fewer sensors to cover a large area. This is overall very early in development still. We've detected, we've proven we can detect and provide a precise location. The challenge here now is on the government side to respond. You've got to respond in a boat. Do you have a drone go? There's all sorts of things we're working out.
The key area we're in right now is called Sabah. It's a state of Malaysia, and it's one of the richest coral reefs, which is being progressively destroyed by this practice that the fishermen have. Possible commercialization, if the government will participate in expanding this along more of the shoreline. I'm going to finish up with another part of ShotSpotter Labs, which is addressing this consumer business opportunity. Is there an opportunity there? We engaged some candidates for master's degrees at Stanford, locally here, and they just completed the first phase of a project for us, was to see what was the opportunity, what would it look like, what would the business model be. We thought we'd just share that with you quickly.
The idea that they came up was take the data that we have about gunfire incidents and work with companies that do mapping applications or ride-sharing services, where consumers are traveling or somebody's driving somewhere. If there's a gunfire incident, to reroute them for safety purposes. There's some interesting benefits to the companies that are involved in this. It seems like an opportunity that if we were to move forward with it, would take little development on our side. We have an API that could automatically feed organizations that have these kinds of products, very precise information about where these incidents are, and they can use their technology to reroute your driving so you don't run into one of these incidents.
This may or may not come to full commercial deployment, but I think it's an interesting idea that we're just kind of incubating for the company now. That's it. On to the technology piece.
All right.
Thank you, Sam. My name is Paul Ames. I head up the technology team for ShotSpotter. Also on the line from his home is Rob Calhoun. Ralph already mentioned that Rob Calhoun is one of the founders of ShotSpotter. He's still working here. He's part of the technology team and a key contributor. He actually heads up the algorithms team that deals with the core location and detection algorithms that make ShotSpotter actually work. We'll be hearing from Rob a little bit later. I think you already know this. ShotSpotter is a wide area, acoustic gunshot detection and location and alerting system. That's a pretty long phrase. ShotSpotter was architected back 20 years ago as a technology that provides a precise location, very low false positives and very low false negatives across any size coverage area.
We're comfortable in a small city or in a security deployment, and we're also comfortable in a large city like New York City or Chicago. ShotSpotter is unique. I mean, the technology is unique in that not only do we do the core detection and location task of gunshot detection, but we also deliver a very sophisticated suite of applications that provide real-time access to gunshot notifications. Sam, I think, or maybe Ralph was talking about publishing a gunshot within 30 to 45 seconds of a trigger pull. We do that. We also provide applications that provide historical information for a customer in their gunshot area, and that's very useful for crime analysts or for detectives. What was I going to say? I'll come back to the key points in a second. Today what we're going to do is we're going to cover three separate areas.
One is we're going to focus on the core detection and location technology and do a bit of a deep dive. Ralph gave a little bit of a warning, it may get a little technical, but it will give you a sense of what we're about. We're going to look at go up the technology stack and take a close look at a couple of applications. We'll finish off comparing ShotSpotter and the way that ShotSpotter is architected with competitive systems. The point that I was going to make that I've forgotten was ShotSpotter has been around for 20+ years. Rob's been here from the beginning. During that time, we have plowed a lot of money into developing the system, and we've also learned a lot. The quality of the system and the barrier to entry is extremely high.
Let's take a look at the core detection and location technology. When I think about gunshot detection, it's a problem of that's somewhat analogous to what Sam was talking about, where with a sales funnel, you have an almost infinite amount of potential customers out there, and you've got to work them through a sales funnel until you get a closed sale at the bottom. For us, it's a little different. What we're doing with gunshot detection is we're listening to all the sounds that occur in a coverage area in a city. There's pretty much an infinite amount of sounds that occur. Working through the funnel at the bottom of the funnel are qualified gunshots. It's kind of a filtering process that gets you to the bottom.
For us, the stages that you go through for gunshot detection, I like to think of it as a four-stage process. The process starts with the sensor. When we deploy a square mile of ShotSpotter coverage, we install roughly 20 to 25 sensors in the coverage area, and their goal in life is to listen for impulsive gunshot-like sounds. When they hear an impulsive gunshot-like sound, they take the waveform, they translate it into a set of numbers. They take the precise time of arrival of the sound at the sensor, and that's used for location a little later on, package up that information, and put it into what we call a pulse packet. The pulse packet is sent over the private network, delivered by the cellular carrier to our AWS-deployed back end that does all the heavy lifting. That's the first phase.
What you've effectively done at this point is you've taken this infinite amount of sound, and you've qualified those sounds down to impulses, gunshot-like sounds, that could be a gunshot. The second phase of this filtering process is performed by the software in the cloud that is enabled by the sparse array of sensors deployed in the city. There is a key attribute of a sparse array that we refer to as a spatial filter, and the way that a spatial filter works is, if a loud sound occurs in the community and multiple sensors in the sparse array hear that sound, by definition, the sound was loud and gunshot-like. Conversely, if an individual sensor hears a sound that the other sensors didn't hear, you know it's quiet, and it's not likely to be a gunshot.
Counterintuitively, if I'm a sensor in a sparse sensor array deployed to a city, most times I hear a gunshot off in the distance. You're not actually listening for a loud sound, you're actually listening for a pretty quiet sound. When we're able to figure out that multiple sensors have heard the same sound, we're able to calculate a location. We create an incident entry into our database, we reach out to the sensors, grab some audio of the event, and we move to the next stage. What we've done at this stage is we've taken all these impulsive sounds, we've winnowed them down to an impulsive sound that was loud enough to be heard by a set of sensors in the sparse array. We've kind of gone down the filter. The third stage of the filtering process is machine classification.
We gather up all the data that has been collected to date, we run it through a machine classifier that we'll talk a little bit about in a second. The goal here is twofold. One is to filter out incidents that are classified as non-gunshot. For example, we may wish to take incidents like a helicopter incident or a firework and say, "You know what? Let's not push it further down the funnel. Let's suppress it." The other goal of the machine classification is to provide the human reviewer the assessment of the sound so that they can then use that as a baseline for their human classification.
Again, this third phase has said, "Okay, I know we found a very loud sound that was heard by multiple sensors." We've now filtered out a bunch of incidents that were, for example, helicopters, and we then move to the fourth and final phase, which is human review. The role of the human reviewer is to take that incident, listen to the audio, listen to the machine classification, and other data that's been gathered along the way, and form the final determination as to whether or not something is a gunshot. If it is a gunshot, they press the publish button, and within 30-45 seconds, that incident is delivered to a police officer in the field. If not, then the thing gets thrown away.
Again, just to review, it's kind of a filtering process, the infinite amounts of sounds in the city through to the point where we've got qualified gunshots. To achieve this, there are a bunch of technical problems that we need to solve, and what we're going to do is just talk fairly quickly about three technical problems in a little bit more detail. The first is, how can we reliably hear a gunshot that's off in the distance in a noisy city where the gunshot is only just above the sound of the city? Number two is we're going to look at how do we figure out from all these sounds that we hear when the sounds refer to the same point source. How do we actually do that correlation?
Third, how can we do a really bang-up job on machine classification? I'm going to hand it over to Rob Calhoun, who should be on the line. Rob.
Hi, my name is Robert Calhoun. I'm one of ShotSpotter's founders. I'm going to discuss the adaptive noise filtering algorithm that we use on our sensors. Gunshots are simple impulses, so their energy is spread over a broad range of frequencies. This first slide here compares the theoretical power of a gunshot in blue with that of an actual gunshot recorded at a distance of 170 meters. That's in red. Because of the filtering effects of the local environment, the measured power spectrum is more complex than the theoretical power spectrum. In fact, the received signal is more characteristic of the environment in which a firearm is discharged than it is of the firearm itself. For this reason, the sensors must accept a wide range of impulsive sounds as plausible gunshot impulses.
A hammer and a nail fixing some clapboard at 50 meters sounds very much like a gunshot at 500 meters. Competing with the detection of gunshot impulses is unwanted noise. The acoustic background noise in urban areas is high and continuously changing. Rush hour traffic comes and goes. HVAC systems cycle on and off. Helicopters fly overhead, and so forth. In order to optimize the detection range of ShotSpotter sensors, we use an adaptive filtering algorithm. This removes some of the gunshot signal, but it improves sensitivity by improving the signal-to-noise ratio. We need to improve the signal-to-noise ratio because we must detect small, inexpensive handguns, as well as high-powered ones. The amount of propellant in a handgun cartridge varies by an order of magnitude depending on the type of ammunition.
To detect low-power firearms in built-up areas, we need to be able to detect signals with a very low signal-to-noise ratio. Next slide, please. This slide shows the results of ShotSpotter's adaptive filtering algorithm on the received signal picked up by two different sensors from the same six-shot gunshot incident. In the top left, in red, we see a waveform, that is, the signal amplitude as a function of time, detected by a sensor that is 285 meters from the shooter. The six shots are easy for us to see in the waveform. Similarly, the onboard pulse detection algorithms have no trouble triggering on these impulses. In the lower-left corner, we see a more difficult case. This is in blue. This is a signal from the same incident received by a sensor that is 570 meters from the shooter.
The signal is fairly strong, but it is obscured by background noise. Since our signal is broadly distributed in the frequency domain and the background noise is always changing, we use an adaptive filter to maximize sensitivity. The filter continuously updates a model of the background noise, then boosts or attenuates each frequency band in order to optimize the sensor's sensitivity to gunfire. The time constant for the background noise optimization is around 15 seconds. The center column shows the filters that were in use during the processing of these two signals. The background noise in the top signal is mostly from a nearby highway because this incident was during rush hour. This filter acts like a simple high-pass filter and minimizes the traffic noise. The bottom signal contains traffic noise, machinery noise, and wind noise.
Wind noise is a significant problem for us because our sensors are usually mounted up on the top of buildings to keep them away from street level. It includes some high power peaks at specific bands. The adaptive filters put notches in the filter to minimize the effect of these noises on sensitivity. The deep notch there is a -12 dB of attenuation applied to that very narrow frequency band. All the filtering is done in the frequency domain to avoid phase shifts, which are undesirable when measuring time of arrival. You can see the effect of this filter in the right column. After adaptive filtering, both sensors triggered on all six shots, which is what we want. Next slide, please. Next, I'm going to talk about pulse set selection. Triggering generates a pulse packet that is sent to ShotSpotter's data center for location processing.
Location processing comprises two steps. The first is determining the set of pulses that should be used in multilateration, and the second is computing the location from the time difference of arrival, which is what multilateration is. Now, computing the location from a set of time diff of arrival is a fairly straightforward mathematical algorithm, and it has been known for many years. It was even used in World War I. The first of these problems is actually the more difficult one. Which set of time difference of arrivals are we going to use for multilateration? The pulse selection problem is NP-complete, which is a computer science term that means we shouldn't expect to find an efficient algorithm for determining the optimal solution. Next slide, please.
Since we know not to expect one algorithm that is the best for all possible inputs, ShotSpotter uses multiple pulse selection algorithms that run at the same time. Each solution gets scored, and we use the best-scoring algorithm. That way, the algorithm that wins tends to be the one that's most suited to the particular situation that it encounters. These pulse selection algorithms include a combinatorial optimizer. It's not that we're trying all N choose K combinations because there are too many of those, but we choose subsets that are likely to be due to acoustic point sources, and we combine it with greedy algorithms that ensure we collect all of the relevant data associated with a single shot. We also do cross-sensor pattern matching, which I'm going to talk about a little bit more, and some forward solutions.
With all of these, the goal is to take an incoming stream of pulses that comprises gunshot pulses, echoes of gunshot pulses, and non-gunshot impulsive noises, construction noise, helicopters, all kinds of noises. identifies those subsets that are consistent with acoustically generated point source. Next slide, please. This slide shows how ShotSpotter's pattern matching routines function. The diagram on the left shows the acoustic signal measured by the sensor. This signal is not available to the algorithms running in our data center, and we don't need it. Instead, the data center creates what we call a synthetic time domain representation of the gunshot data that's built up from the pulse packets that were transmitted to the data center. The weight assigned to each Gaussian peak is associated with the likelihood that that pulse is a gunshot.
We use the mathematical techniques of autocorrelation and cross-correlation to align the shots across multiple sensors. The pattern matching locks in on invariant shot spacings. This is a highly effective way of ignoring the echoes that are present in most urban locations. Echoes are very common, but no two sensors share the same path that the echoes travel from the shooter to the sensor. Therefore, the echo time delay varies from one sensor to another, while the shot spacing between shots is just determined by how fast the trigger is pulled. That's a great way of aligning all the shots and associating all the pulses from shot one together without getting confused by the echoes. This technique does have one disadvantage, which is that it does not work well when the shooter is moving. That's more common than you would think.
People drive their cars, and they shoot out from the cars, and then they drive away. Say a shooter is driving towards sensor A and away from sensor B while discharging a firearm. Similar to the Doppler effect, the apparent shot spacing between shots is reduced on sensor A and increased on sensor B. Since the shot spacing is no longer invariant, we no longer have a good pattern match. We handle moving shooters primarily with a different algorithm. Next slide, please. The next algorithm that I'm going to talk about is a forward solution, and this is a computationally intensive approach in which you build a model of the shooter's position and the shooter's velocity, and then you just try and see if it's right.
You pick a location and say, "Well, what if the shooter were here and driving north at 10 meters per second?" You see how well that makes all the data make sense. If it doesn't seem to have a great fit, you move it a little bit and try again. That's expensive computationally, with modern processors, these calculations can be done in near real-time using numerical optimization techniques. What's nice about the forward solution is it can take the motion of the shooter into account, so you can end up with a good pattern match even when the shooter is moving. We don't actually take that solution and feed it directly to the output routines. Instead, we use this forward solution to pick the pulse sets. We take those pulse sets and apply the multilateration routine on it.
That way, the final location is determined by multilateration. If they're starting to go around the corner, we can see that as they start to go around the corner, and we can pass that information on to law enforcement. Paul Ames will be talking about the next step, which is our machine classification system. That's all I have for today. Thank you very much for your time and for your support of ShotSpotter.
Thank you, Rob. I'll be doing a test on that later, of course. Machine classification is incredibly important to us as a company. I mentioned it before, the driver is twofold. One is filtering out stuff that we really don't want to give to humans, and two is increasing the quality of the machine recommendation that we give to humans. Both of these are designed to raise the quality of the end result. Our 34th patent that was issued earlier this year is incredibly interesting. It leverages the advances that have been made in machine learning, specifically deep neural networks, for recognizing objects in either video or in photos. This is the type of technology that figures out that your camera viewfinder has a person in it, or it's the technology that is in your photo collection software that recognizes pictures of your mom.
It's the exact same stuff. What we're doing, what this patent does is it creates what we call an image mosaic. An image mosaic is literally just an image, and the mosaic is made up of individual tiles that describe visually different attributes of the gunshot. We give that image to a machine classifier and train it, and we're able to differentiate between helicopters, between fireworks, and between gunshots. One key thing that I need to call out is we're not only taking information from the waveform, we're also taking information from the sparse array. There's a lot of data that's used in the assessment of what that sound is. We're using this technology today. We've been using it for the last 12 to 18 months in our production systems, and we're using it today to suppress helicopters, as I was saying earlier.
As of tomorrow, we're going to take the confidence information that the classifier gives us and use that to prioritize incidents that are given to the human reviewers. The goal there is, if we feel pretty confident that it's a gunshot, get it to the reviewer quicker. If it's something that doesn't look as though it's a gunshot, then it's not as important, and it can hang around for a bit. We're also going to use this confidence to also publish certain gunshot events. The benefit there for the first responders is getting that notification into the hands of the first responder maybe 20 seconds before, if it has gone through a human review. Incredibly interesting and incredibly important to us as a company to continue to add square miles without scaling up the people within the review center.
We're through the most dense part of the presentation. We'll take a quick look at applications. Q2 of this year, we released an internal tool that's called Review. Review is used by the human reviewers within the Incident Review Center. Its goal is to allow the reviewer to make that assessment as to whether or not it's gunshot or non-gunshot. I'll call out one particular feature of the application. The application surfaces what we call a proximal incident. A proximal incident is particularly useful when there is a repetitive sound coming from a city. I think Rob used the example of a building site where somebody's using a nail gun or somebody's using a hammer.
In that particular situation, it's very valuable for the human reviewer, actually also the machine classification, to see the incident in the immediate area and also to be able to look back in time. That information will help inform a better decision on the classification. In addition to the features that we've already rolled out, we're working with Nasim's team, she'll talk about this a little bit later, to add peer review. In certain situations, having a second opinion, a second human opinion, is really valuable as to whether or not something is a gunshot. Looking to next year, we'll be enforcing policy actually within the application. One of the use cases here is, in our security applications, when a gunshot is published, they're pretty infrequent. We actually also pick up the phone and call the security desk, for example, within the school.
What we'll be doing is building into the application that capability to make that call. The second internal application that I'll just touch on is a thing called Dashboard. Dashboard provides complete visibility into all of the workflows, the software that's deployed into our cloud. It's used functionally, cross-functionally, across the organization by customer support, by technology, by customer success. It answers questions. For example, if I'm a product manager, a bug has been reported on a particular version of software on a particular OS version, then one thing that you want to know is how many customers are impacted. We can surface that information through Dashboard. We can also surface information relating to the quality of the machine classification versus the human classification.
For example, is it 50% of the time that a human disagrees with a machine classifier, or is it 10% of the time? Those things are particularly significant. Incredibly valuable tool for efficiently running the business. Let me turn quickly to two externally or customer-facing applications, Respond and Dispatch. Respond is the application that's used by first responders out in the field. The application receives real-time notifications. It puts a dot on a map, and it provides the dispatch address where they should go to deal with this incident. Dispatch is an application that's used by the communication center or often called a 911 center. The goal there is to act effectively as a proxy for somebody calling up. As Ralph was calling it's a digital 911, effectively. This is the tool that they use to dispatch a patrol officer to the correct location.
On a technology level, we're not a large company. How we develop our applications is critically important. We need to be able to do it extremely efficiently. We go to a lot of effort to ensure that our applications are as much as possible, leverage a single code base. We use containers to deliver applications to your iOS device or your Android device or a Windows machine. This is incredibly important for us to maintain a high velocity of enhancements to our applications. In 2019, we added the capability to on-demand request an Investigative Lead Summary through the application. Nasim's going to talk about this, and a couple of other features that have been particularly well-received.
Looking to next year, two features here listed are focused on situations where our view of what gunshots have occurred in the community is a little different than the police department's view. We're working on a process of reconciling those two views so that we can have a conversation about performance of the system. Moving on to competitive architecture. When you survey the competitive landscape, solutions generally fall into two categories. The first category relies on a sensor technology that is generally called a proximity sensor. The other category is a multi-mic cluster sensor. A proximity sensor is typically offered as an add-on to a camera, or it could be an add-on to a smart streetlight. Generally speaking, it relies on a single microphone.
What it's doing is that it's providing or it's committing to detect gunshots that occur in a small radius around where the sensor is. The radius is generally somewhere between 100 meters and 200 meters. With one microphone, this sensor is unable to precisely locate the gunshot. Literally, such a sensor is able to say, "Yeah, I think I heard a gunshot, and it's somewhere in this area." The major challenge that any vendor has with developing a proximity sensor is with false positives. False positives are particularly problematic for a police department. If you dispatch a patrol officer to a location and it's proven that there is no gunshot, over time, the first responders lose confidence in the system and start to not use it. If we look back in our history as a company, prior to 2011, where ShotSpotter was deployed as a customer premises deployed system.
In that era, it was up to the dispatchers to actually make the assessment as to whether or not the incident that was reported was in fact a gunshot. Unsurprisingly, the quality of the classification was really low, and the solution at that time gained a reputation as being a false positive machine. False positives are really problematic. Moving on to multi-mic cluster sensors. Multi-mic cluster sensors came out of the military applications. Typically, these solutions have three or more microphones in one plane, and one microphone off plane. The reason for this configuration is it's the same theory that humans have two ears. We have two ears so that we can figure out the direction of a sound.
With four microphones and one off plane, the sensor is able to say, "The gunshot came from here, or it came from here, or it came from here." The capabilities of a multi-mic sensor depend on whether or not the gun that was fired was supersonic or subsonic. In the case of supersonic, so an example here is a AR-15. When you shoot an AR-15, the bullet actually travels faster than the speed of sound. In that situation, the sensor is able to pick up the shockwave from the bullet, and from that do a couple of things. It's able to calculate the trajectory of the bullet. It's also, in some situations, able to calculate the caliber of the bullet. Through some clever technology, it's also able to figure out the range of the shooter. Through that, you're able to come up with a pretty decent location.
The bullet must pass within 30 or so meters of the sensor. Effectively, the shooter needs to be shooting at the sensor, or give or take. In a subsonic situation, so a regular handgun, the bullet is subsonic. What the sensor is doing is that it's detecting the muzzle blast in the same way as our sensors detect the muzzle blast. With multiple microphones, the sensor is again able to say, "Yeah, the sound is coming from somewhere over there," but there's not enough information to say, "Here's the range." It literally is somewhere over there. Useful, but not incredibly useful. I've already said this, but I'll kind of underline it. For the supersonic situation, if somebody is shooting at the sensor, it's a great solution. It's fantastic. In a normal city, people are not shooting AR-15s, and they're not shooting at sensors.
In the time that we've deployed ShotSpotter, it's extremely rare to hear a bullet whiz past a sensor. Every now and again it happens, but not that often. What I have here is a comparison of the benefits and the limitations of the different architectures. For ShotSpotter, we do an incredible job at delivering wide-area gunshot detection. The system was built from the ground up to deliver wide-area gunshot detection. It doesn't matter where the gunshot is in that coverage area, we can figure it out, we can detect it, and we can precisely locate it. Because of the sparse array, because of the multiple layers of filtering that I described, we're able to deliver for the customer extremely low false positives and false negatives. Nasim's going to talk about court-admissible evidence, but we're the only system that provides that capability.
The sensor, the technology, the approach that we take is scientifically proven and available to be presented in court. A limitation of our system is that we don't provide caliber. To be honest, it doesn't come up for police departments in the U.S. It's interesting, but it's not important. The summary here is the technology is proven. We've deployed to 100+ cities. The technology, important as it is offered as part of a complete service. Moving on to proximity. There is, I think, out in the marketplace, a perception that this is a low-cost solution. I think this comes from two sources. One is the incremental cost to add proximity detection to a camera generally is not that high. The other is the perception is that gunshots generally are going to occur in this area, this area, and this area.
From our experience, that doesn't happen. It occurs in a very wide area. Although the perception is that we could install a few of these proximity sensors, the reality is to stitch together, let's say, a mile of coverage, it's just not practical. It is actually quite expensive. With proximity sensors, there's a hidden cost there on cost to respond. There's the initial purchase cost, but the cost to respond can be incredibly high. As I described, that particular type of sensor doesn't provide a location. It's just somewhere around here. High false positive rate. Jerry Ratcliffe, this is mentioned at the bottom of the slide is a professor out of Philadelphia. He's an ex-cop from the U.K., from the Metropolitan Police. He did a study of a proximity sensor deployed to Philadelphia, and I would encourage you guys to take a look at this study.
It supports what I've been saying about the number of false positives. Keep in mind when you read this paper that Jerry tars all gunshot detection architectures with the same brush. If you read it closely, he's actually talking about one specific solution that is a proximity sensor. Moving on to multi-mic. For a supersonic bullet, extremely good. Awesome solution. For subsonic, very limited coverage, poor locations. The summary there is, if you are deploying a multi-mic sensor for a military application, or if you're deploying it for perimeter protection, it's a very good solution. For subsonic, or for using that type of sensor in a public safety environment, it is not a good fit. If you look at the websites of some of the vendors out there, over the last perhaps year, some of the vendors have started talking about collaborating sensors.
If I was the head of technology for one of these companies, I would be doing the same thing. The problem here is, or the observation I'll make is, what they're doing is they're starting out the learning curve that we've been climbing for the last 20 years. That's not to say that it is actually going to take them 20 years to get to the same level of performance, but it's going to take a lot of time, and it's going to take a lot of investment. By the way, they're going to have to keep a close eye on our pretty extensive patent portfolio. I wish them well. I think I'm done. Handing over to Joe. I don't think the microphone is on. Is the microphone not?
Does the microphone not work here?
Very good. It's working.
Okay, great. Good afternoon, everybody. My name is Joe Hawkins. I have the privilege of heading up ShotSpotter's Operations Department. I'm in the middle of my eighth year with the company, and I can tell you that having been in the technology business for a long time, and I have been for a long time, this is by far the most inspiring and gratifying place I've ever worked. It starts with the great purpose that Ralph talked about at the top of our afternoon together, and really ends with the incredible team of people I have the privilege of working with. As cool as the technology is, that Paul and Rob just described to you, for it to make an impact in solving the gun violence problem, you got to take it for a walk in the real world.
You got to deploy it in the cities where gun violence occurs. That's what I'd like to describe for you today. I'm going to go fairly quickly to try to catch up a little bit. Operations generally has two major missions. First is we lead the effort, the company-wide effort, to implement new service for our customers, whether it's public safety or security-focused, and get that service up and running. Secondly, we manage the installation, remote monitoring and management, and when necessary, repair of the distributed network of acoustic gunshot sensors that live all across the country, wherever they're deployed all across the world.
It's a service assurance job that we take very seriously to make sure that not only is our service working as best as it can be on day one, but it continues to work optimally for the entire existence of ShotSpotter in every city where we operate. I'm going to be spending most of my time talking to you about the implementation side, not so much the service assurance side. We have a project management team that leads these implementation efforts, and they wear a couple of hats. One is very common to project management in any business, and that is they lead project teams, interdisciplinary cross-functional project teams, to execute a plan to achieve the objective of delivering and implementing that new service, on schedule, on budget, delivering the expected quality of service.
They are the face of the company, the primary contact with a customer to communicate what those plans are, to keep them informed of progress, and when necessary, to get what we need from the customer to help make that a collaborative effort. They also lead the internal company's team, that includes all walks of professional contribution people, functional roles within the business, who are delivering their parts to make the whole thing come together on schedule, as I said, in a very predictable fashion. At any given time, we are running multiple concurrent projects, many at one time, and each of our project managers also is typically managing more than one project at a time.
The second hat that they wear is very unique to our business, and that is they are also the designers of the sensor arrays that we build and manage in all the cities, including picking where sensors are going to go, negotiating for permissions to get those sensors installed, to make sure that the backbone of the service is in place. They also plan and coordinate a live-fire test with our customers, whereby at the very end of the implementation phase, we go out-Pick a handful of locations within the coverage area, and the customer, the police department, fires a large number of live rounds into a bullet trap. You can see a little picture there on the lower right as an example of that, so that we can get some good empirical data about the baseline performance of our service as it's going live.
We will occasionally repeat those tests over the course of perhaps years later as we make changes to the network or as the city may change, to see how we're doing. It's an interesting thing that we do as well. Lastly, we make sure that not only as we are getting the customer to sign off and accept service to begin a service, and we call it going live, that we've done a launch readiness review internally with all members of the business that will be delivering the service to the customer to make sure that we're ready to deliver as well. A little bit about the project management team. This is very unique. We're the only company in the world that has done what we do as much as we have. We've got a seven-person project management team.
You see those gentlemen up there in the upper right-hand corner. Collectively, over 40 years of experience designing and implementing ShotSpotter service. They've got backgrounds before ShotSpotter in all kinds of walks of life, technical, telecommunications, even sales, with pretty good public safety backgrounds as well, and they've all got the ability to establish and forge deep relationships with law enforcement. They communicate very effectively with our customers, and they understand where our customers are coming from and how we are helping solve problems for them. In terms of our body of work, at this point, we have implemented over 150 projects deploying ShotSpotter services, and almost double that many discrete, separate ShotSpotter systems. Some cities may have more than one system, and in fact, several do. Over 700 sq mi of service. It's a pretty big footprint.
Within those implementation efforts, we've seen projects with all kinds of diversity, both in terms of size and in terms of the urban environments in which they operate. 10% of those projects have been 10 sq mi or more at a bite. Another 20% between 5 sq mi and 10 sq mi. Most of our projects have been three square miles or less in terms of size, and that was actually fairly common in the earlier days of our business. In the last couple of years, it's been increasingly more common that the first implementation project will be of the medium size to larger size. But across all those projects, we've seen just about every type of urban environment, suburban environment, even semi-rural in some cases, and have learned a great deal from that experience.
We're able to do things and understand things now that we wouldn't have in year one or year two of our business. On the security side, we've now deployed nine college campuses, the largest being three and a quarter square miles of the campus and the surrounding community, where the staff, the students, the faculty work, shop, dine, and so on and so forth. We've implemented one freeway security network that I think most of you have heard about. This is the Contra Costa Freeway here in the Bay Area. We have installed and operate one secure government facility, which is something that's called a critical infrastructure facility. For reasons of confidentiality, we're not able to talk about who that is, but they obviously take their security very seriously.
When we look at a project, whether we're talking about the first one or the 200th one, we now have a model that's pretty repeatable and pretty well-known to us. The overall project, if you look at it in the context of the entire customer relationship, which starts when sales is just beginning their prospecting with marketing and goes on for years of an active relationship, the project for us is a discrete point in time that starts when a contract for a new service or expanded service is signed and ends when the service is activated, or we go live. It's that middle part that is the project manager's domain and consists primarily of implementing the technology and onboarding the customer. The pie chart you see there just represents the relative size of the major parts of the work effort.
The project manager does some of these things themselves, principally plan and design, permissions acquisition, and the live fire test. The other major parts of that wheel are done by other members of the project team working with the project manager. The two biggest in terms of energy, as you can see, are permissions acquisition and customer onboarding. Paul Reeves will be talking a lot about the customer onboarding practice when he gets up here to speak with you. Now, when you overlay that same pie chart over time and look at a schedule, you'll see a couple of things of interest. One is, a lot of this work is done concurrently or in parallel. It's a multidisciplinary, concurrent engineering effort. The long pole in the tent that defines the critical path is the permissions acquisition aspect of the project.
All of our sensors have to go on properties, whether they be buildings or whether they be streetlights of the sort, that are owned by third parties that are not us and not our customers, which makes it a sales job, basically, to go and ask for permission to put sensors on a property where they're not getting compensated. As fast as the permissions acquisition process goes, the faster our project goes. Lastly, you'll see a couple of control gates. These are these major tasks that are highlighted in yellow. One at the very beginning of the project, one just before we go live, and one in the middle of the project, which we call control gates, where we really circle the wagons internally, see what we've got coming ahead of us, make sure we're aligned on the plan.
As we go through the project, we've learned a lot of things. We refine those plans, make sure we stay in sync to deliver what's expected without surprises. Our project teams also look very similar from one effort to the next. Project manager, I've already talked about their roles and responsibilities. As I mentioned, Paul Reeves will be talking to you about customer success with the big lift of onboarding our new customers. We've got contributions from network operations dealing with the provisioning and configuration of the back-end cloud services. We have a field service team with our own employees and a network of third-party field service contractors that deal with the installation of sensors in the field. They're supported by a supply chain team that deals with the contract manufacturing, purchasing, and delivery of materials where and when they're needed in the field.
Lastly, but not leastly, a customer support team that deals with integration of ShotSpotter technology with third-party systems like video management systems, CAD, records management systems, and so forth. They also support our live fire tests in real time and post-test analytics, and make sure that they're ready to start delivering service when we flip the proverbial switch. I want to close just by talking a little bit more about sensor array design, because it is a bit of our secret sauce. It's something that we refined our own practices and skills at over years of practice. The first thing to understand is there are a couple of guiding principles that inform our designs. The first is that what we build and implement must meet or exceed our service level commitments for gunshot detection and location accuracy. That's the SLA+ .
Our goal is always to do better than the 90% that Ralph referenced earlier today. The second principle is a little bit more nuanced, that is the principle of overbuilding. That means, in short, that we're putting a lot more sensors into the city, into the community than we need to achieve that SLA, we do it for a couple of reasons. R&R stands for redundancy and resiliency. I mentioned that we're putting these sensors on properties that we don't own, our customer doesn't own. Again, a lot of times, they're in communities where it is not uncommon for a building or a business to close and the lights go out. A home is sold and the lights go out. The BGE bill or the electric bill just doesn't get paid, the lights go out, and we lose a sensor temporarily.
We don't always have immediate access to those buildings. We anticipate that any given point in time, there's going to be some number of sensors in our coverage area that are not online, and it can affect our service. We put a lot more than we need out there knowing that some of them are going to be off at any given point in time. That allows me to manage my field service and repair business in a conservative, gradual, thoughtful way. As part of that, I'm also cost shifting, meaning I'm over-investing upfront where I can capitalize the equipment and labor that goes into building out the network and amortize that over several years. Thereby not have to be incurring very expensive, unpredictable, lumpy maintenance costs by having to respond to every downed sensor within 24 hours.
That actually turns out to be a wonderful lever to being able to ensure that we've got predictable, reliable operating costs, which helps our margins, while also maintaining very predictably high quality of service. Those are guiding principles. In terms of the practices, this is a blend of art and science. We're all very well knowledgeable about the speed of sound, the physics of acoustic propagation in an urban environment. A lot of what we do is the art of understanding how one city is different from the other. In fact, every city is unique in terms of how it sounds. We take environmental factors into consideration. We have to look at geography. We have to look at topography, the size of roads, the amount of traffic, how big the buildings are, how densely packed they are, what they're made out of, man-made structures, bridges, viaducts.
We've got weather. Is there prevailing winds? That's going to mean there's a wind grade always coming from the northwest we can compensate for by how we deploy sensors and so on and so forth. Environmental factors and how that affects sound propagation is an important consideration in deciding how many sensors we need in which place. In placements, picking the buildings or the sites where we put sensors have to be both suitable and mask-friendly. By that, I mean suitable, they need to be accessible to us. Ideally, they're going to be up high where they can hear to the horizon. They can have a good proverbial line of sight to sound at great distance without that sound being attenuated a great deal. Mask-friendly meaning we don't like neat, clean grids.
We like things to be asymmetrical so that wherever the gunshot occurs, you're going to get that good time difference of arrival advantage that helps with the multilateration problem-solving later on. Again, sometimes we go very dense when acoustic challenges are pretty heavy, and we can go pretty light with our sensor array density when sound is going to propagate very cleanly and smoothly and predictably at great distance without a lot of interruption or attenuation. We employ things called hints or exclusions. You could think of these as geo-sensors where I can mask or suppress sounds from certain parts of the coverage area where I know that I don't want to be hearing stuff. Classic examples would be the outdoor police gun range, where they're shooting all the time. Sometimes these are in our coverage areas.
We can put a geo-fence around that gun range and effectively suppress all the gunfire that's coming from inside there because we don't need to hear it. The police doesn't want us to alert them to it. We can do the same thing if there's a major construction effort going on in a city where they're going to say, "Hey, this whole quarter mile is going to be demolished. There's going to be bulldozers, wrecking balls, jackhammers." We can geo-fence that out temporarily until the construction's done, then turn it back on. We use a lot of tools, our project managers do.
In the lower right, you see a kernel density heat map that informs us about relative sensor density, and we overlay what we're looking for in terms of density or sparse parts of the array based on what we've already discovered in terms of the environment that we're presented with in the city. We use both remote surveys, tools like Google Earth, physical site surveys to make sure buildings are what we saw on Google Earth, and most importantly, we can hear the sound of the town, drive the different parts. We look for commercial or industrial sites that produce a great deal of impulsive noise where we may want to compensate by greater sensor density. We do cellular surveys to check for latency and speed of the cellular networks that we're relying on.
One of our most important tools now that's gotten better over the years is our own experience in deploying ShotSpotter services to different cities. Any city we're going to go into now, as much as it's going to be unique in some ways, is going to resemble a lot of other cities that we're already in and have a tremendous amount of empirical data about gunshot detection and the performance of our technology infrastructure to support that. That helps us get smarter about how many sensors we deploy optimally so that we're not underspending or overspending to get the best level of service we can. That tool set is improving every time we deploy a new system. Lastly, there are practical considerations and real-world constraints we have to deal with.
Sometimes we may really want that 11-story building right in the middle of town that's the tallest one around. We're just not going to get their cooperation. We'll have to go with plan B, which might mean two or three locations that are our second-best choice around that building. We always find a way to get what we need to make our service work. Those real-world constraints do drive us in certain directions from time to time. Last thing I just wanted to show you is a couple of typical examples. These are two coverage areas that actually exist. They are both three square miles in size. The one on the left you see is a typical California Central Valley city. Pretty residential, very flat, typically single-family homes, detached residences, one story, maybe two stories.
Pretty wide streets, not a lot of traffic, not a lot of industrial noise. We're able to cover this very, very effectively with 17 sensors per square mile. On the right, you see a typical northeastern urban city, much more densely populated with people, with buildings that range from three to five-story brick and stone structures to mid-rises and high-rises. You've got commercial districts, you've got restaurants, you've got industrial noise of all kind, trolleys and streetcars and trains. It's a lot noisier. We get similar performance there, but at 50% more sensors, 25 sensors per square mile, all getting the same result. These are two examples. There are lots of other ones that would look like these or might look slightly different.
You might consider them to be prototypes that help us know what we're going to do in this city that we're going to tomorrow and the city we're going to go to next week. It's a very cool and exciting and rewarding job, and that's your quick tour of ShotSpotter project management and service implementation. Now I will give you Nasim.
Real quick, we were going to have a break, but because we're running a little bit behind, we're going to just power through. If you do need to take a break, please feel free to excuse yourself. We're going to keep going.
All right. Thank you for bearing with me. I know it has been a long day. My name is Nasim Golzadeh, and I'm in charge of customer support and professional services for ShotSpotter. I was very fortunate. I joined ShotSpotter in March of this year. Prior to that, I was with CentralSquare Technologies, one of the leaders in public safety technology, and I was in charge of operations and professional services for them. Let's take a look at professional services organization. Our functions are organized in five different groups or different functions. I would like to take this opportunity to just probably just share a few of the highlights with you. Let me begin with the Incident Review Center or IRC. The reason is that that is at the core of our services. That is the center that never sleeps.
They are always manned and operational 24/7, year-round. Their function is the human review of the incidents that we detect. Also, they provide tier 1 support to our customers. Let us take just a few minutes and review that human review process. Paul talked about the first factor of that classification, which is performed by the machine. It's really sophisticated, complex, and outcome of that is a classification, which means gunshot or non-gunshot, and a confidence level. That goes to our highly trained reviewers, and they use that information along with the audio from the incident, the waveform, the way and the pattern in which the sensors participated in the detection, and also other information. For example, proximal incidents. For example, the season, the time of day, the location. They use all of that, and ultimately, they come up with a final classification.
Either it is missed or it is a gunshot, and they publish. When I talk about that, it takes several minutes. In actuality, it happens always under 60 seconds, and the average is closer to 40 to 45 seconds. When in doubt, they take advantage of peer review. Right now, that is a manual process, but shortly, within the next few days, we are going to move to a more automated process to have more discipline and make that more efficient. Also, we try to provide situational awareness to the officers whenever it is possible. For example, when there are multiple shooters or it is an automatic weapon, or it is a high-capacity weapon, so on and so forth. IRC is a service organization, so it is very important to us to be efficient and to be scalable.
We pursue that through many different avenues and ways, and one of them is being really smart and optimizing our staffing. We have performed some studies, looking back at historical data back 18 months, and we have identified the patterns in which incidents come into the IRC, and those are very much peak time driven. They differ vastly based on time of day and day of the week, and that is for normal weeks. The fluctuation between the lowest volume to the highest is about 500%, and that is exactly how we man IRC. At different times, you go to the IRC, you may see two reviewers, sometimes you see 10. Of course, during just ultra peak times such as 4th of July, Memorial Weekend, and so forth, that staffing could be much higher than that.
Also we've worked with our technology team to partner with them, collaborate with them to work smarter and not harder. Some of the examples include our suppression. At any given month, we suppress more than 65,000 helicopter incidents at almost 100% accuracy. That is very meaningful. Other examples, Paul mentioned the prioritization. Today, we prioritize incidents for review based on whether they are probable to be gunshots or not gunshots. Within a few more days, we're going to make that confidence-based, which is even smarter than that. There are numerous other examples, workflow optimization, and so forth, but that is a very continuous process. We provide feedback to technology team, and they are very responsive to make IRC more efficient and more accurate. Talking about accuracy, that is obviously probably the most important thing in terms of having credibility in this space.
Our data shows that even though between 2018 and 2019, the number of gunshots published has increased by 34% because of our expansions, extra months, new customers, our accuracy has also increased, and that is very meaningful. Although it was a really tall order to improve that accuracy from 98.8% for our classifications. Now we're higher than 99%, and we strive to maintain and improve that accuracy as much as possible. That is a perfect segue to discussing support because IRC is also in charge of tier 1 support. Unlike many other public safety vendors or technology vendors for that matter, at any point, customers can get a hold of a live person via chat, phone, or email.
Some of the most typical cases that go to IRC for resolution are password reset, sometimes questions about how to use the application, reporting a missed gunshot, misclassification, request for audio, and so on and so forth. Having said that, there are some situations, some support inquiries that are beyond capabilities of IRC because they are more complex or they require more time, and that is where they escalate those to our tier 2, and tier 2 takes more time, and they have more technical capabilities to resolve those issues, and that is a reactive support. That is after an issue has been reported to us. One of the things is for us that we try to be most proactive, and that starts even pre-go live as we partner with our operations team, participating in the live fire testing. We acquire more knowledge about the customer.
During the lifetime of the customer, we continuously monitor the performance, and if we see any indications of degradation or any probable issues, we tend to dive into that and resolve that even before that becomes a crisis. At a really high level, we receive about 1,300 to 1,400 support inquiries every month, and 83% of them are resolved at tier 1. Something that is extremely powerful, I think that is incredible, is that those tier 1 incidents or issue support inquiries, we achieve a 99% real time or first day resolution for those. That is pretty incredible. The 17% that go into tier 2, even though those are more complex, we have about a 63% first day resolution and 85% or more within seven days. One of the unique aspects of our support model is that we have a named actual TSE assigned to every single account.
That allows those people, our TSEs, technical support engineers, to develop knowledge of that customer and build relationships with them. Ultimately, that is a much more effective support model. We talked about our mission. It doesn't end with detecting and alerting to gunshots. We want to reduce gun violence, and that is where our forensic services and litigation support comes to play. We are the only organization that can provide detailed forensic report because that is a court-admissible document. It is created by one of our forensic engineers, and it has really valuable information about the incident, including the precise location and which timing of those shots. It's so accurate that we stand behind it. If needed, we testify to that in court.
One of the things that we did observe, however, was that oftentimes our customers were requesting DFRs, not for court purposes, for investigative purposes. That was not super effective for us from an efficiency perspective and not very timely for them. Because of that, we created another product, and that is the Investigative Lead Summary or ILS. Customers call that a game-changing product because it can be produced directly from their handheld devices, from our Respond application, real time, and it is on demand. Doesn't cost anything, doesn't take any time, and it provides precise location, timing, and sequence of every gunshot. Sometimes it tells a story whether there were multiple shooters or so forth. They find that extremely useful in evidence collection, collecting, showcasing, or conducting interviews.
It's so popular that since it has been launched, on average, we have about 150 of them produced on every single day. We also provide litigation support working with the DA offices. We testify in court of law. So far, we have testified in 17 different states and District of Columbia. We have had successful ruling in multiple Frye and Daubert challenges. Reason for that is our technology is proven, it is objective, it is factual, and it's completely unbiased. Our ShotSpotter evidence has been presented in more than 160 criminal cases across the country. I decided to share this example with you, although it's very sad. It was an unfortunate situation of shooting of an ATF agent. That is in May of 2018.
Based on our DFR, our witness testimony, along with other witnesses, we were able to help in conviction of a known gang member and sentencing of that person to up to 16 years in prison. On right-hand side, you see that is a letter of appreciation from United States Department of Justice, and we are extremely proud of that. Integration services. We do have an API, and that enables us to integrate with other external systems. Those systems are very typical in public safety, such as CAD systems, records management systems, fusion centers, license plate readers, camera systems. That concept is very popular in public safety. Reason for that is that it allows public safety organizations, agencies to take more advantage of those individual systems and technologies that they have in their disposal.
That helps with them to provide more precise, timely, and safer responses to the citizens, and it's more effective for the officers. For us, of course, that is a revenue opportunity, but even more importantly, it's about positioning. When we become part of that public safety ecosystem, then our stickiness increases, and that is an area that we have a tremendous amount of opportunities. So far, we have more than 60 existing point-to-point integrations. We have successfully integrated with more than 25 different systems. In terms of growth, we have grown tremendously, or at least we're on the right track. Last year, for the whole year, we had 10 new implementations, and so far this year, we have more than 25. That is an area that we are really optimistic for, really excited, and we're going to continue to pursue that.
On that note, let me turn it over to Paul. Thank you for your time.
George, we should be applauding. Thank you. You can start the tradition now if you want. Maybe walk-on music. I was going to say next year we should have some walk-on music, but Alan's thinking that's probably optimistic for me. I'm the newest member of the executive team. I run customer success. My name is Paul Reeves. It's been, like, 10 seconds talking about my background because we're running out of time and really focused on the mission of the company. I've been in the Valley for over 20 years working on customer-facing teams. I get into the pattern of getting in early, being a customer-facing leader. About nine years ago, I got into this thing called customer success, which, of course, we all know now what it is.
In the subscription economy, it's about realizing value because we know that people can use software, but are they really getting value from it? I was connected to Ralph via a mutual friend in the customer success community and got really excited about the opportunity because not only was it a company that's doing something interesting and doing something meaningful, which I think we all like to do if we have the opportunity. What's great is this, actually, strategy started several years ago. I may be the newest member of the executive team, and it's gotten more prominent, but the strategy was laid three years ago. The vision or the mission of customer success is all about realizing value for our customers. The reason we can do that is because we have a great strategy, we have a great team, a great product.
This is one of the things I think makes us different. I don't know if you've tried to hire customer success people in the Valley recently. Yeah, right. It's not easy. I've been doing it for years. The profession has blown up in the last few years, I spend a lot of my time trying to find good talent. We have amazing talent because Ralph made the smart decision years ago to have experienced law enforcement professionals who have very distinguished careers for 25, 30 years before they come to ShotSpotter. Our goal in customer success is actually to be the trusted advisor. I don't know how many times you actually felt like you had a trusted advisor when you're working with your customer success professional. From day one, we absolutely have that because our people have been doing this for a very long time.
Many of them actually used ShotSpotter before. I don't know if you know about law enforcement in general. I did not know that, for example, it is common for someone to have a first career of 25, 30 years and then gracefully age out and go find another career to make room for the next generation. We are a top destination for very distinguished, very accomplished people. Like you saw in the video, these people are very sophisticated. From day one, they can make excellent relationships with our customers. That's fantastic, and it's an honor to be working with them. Now, we got a great product, we got a great team, and we have a great design. As you know, onboarding is the single most critical time of the customer relationship, right?
If they're not activated, if they're not using the system, the likelihood that they will remain is extremely low. We're really good at this, as we show in our NPS and survey data. We do this because we're able to talk to them about their goals, their business goals, talk about strategy, and design a program effectively. We don't just get in and start doing user training. We talk about what they're trying to achieve. We can leverage best practice. One of the first things I saw when I showed up, someone proudly walked up to me and said, "Here's our best practice." It was like a 20-pound book. They said, "We give this to customers." I was like, "You give that to customers?" It's a fantastic amount of data, but it's not extremely useful.
What we've discovered is we're going to turn that into a scorecard. That'll be the basis by which we measure the maturity of a given customer. Because we have five or six core practices, we'll be able to turn that into a scorecard to say, "You have this practice. This is where it is today, and these are the KPIs that we're affecting with that practice." Very easily, we'll be able to take that into a change management strategy that we can leverage with every customer. Of course, we are part of the onboarding process. We're with them throughout the engagement, and we're there on go live. That's why we onboard people very effectively. You'll see in our NPS data, people appreciate that tremendously. One of the areas where we're going to grow this year is more around the ongoing engagement with customers.
As you know, you're probably used to being invited to user groups or doing things in an ongoing manner with applications or industries that you like. We haven't taken advantage of that as much. We will. We'll be using the data to find out who's a great customer, making user groups like crime analysts. I just spent the last, a couple weeks ago, five annual reviews. I'm in with a room with the chief, the commander, the patrol, the sergeants, crime analysts. Crime analysts absolutely have a critical role on the team. The chief, the commander looks to the crime analyst, speaks deferentially to their opinions, the data shows that the crime analysts are informing policing policies to have an impact on the communities. We don't have a user group for them. Why don't we have a user group? We will this year.
That's a natural community that can help each other be successful, tell and share their stories. That's one of the things we'll be launching this year. Of course, we're ongoing. We spend a lot of time with our customers, so that's very powerful. You'll see we have an annual account review process, best practice for customer success, of course. Well, we're going to change it slightly insofar as we will continue to do annual reviews. It's part of the process. It has to be. We should be able to do reviews on demand.
One of the last things I did in one of my previous companies was I leveraged modern technology to enable my customer success people to prepare for a meeting in several minutes instead of several days, so that when they actually had the opportunity to talk to brass because brass happened to show up on site, we could be prepared with the story. There's a lot we're going to be able to do to take advantage of modern technology to just be more effective and capitalize on opportunities. We do an annual survey of our customers. We talk about NPS and other key questions to gauge our impact and value to them. I'm going to share those results with you. I think, is everyone here pretty familiar with NPS? You've seen it. Yeah, right. It's part of our company, not just in word, but in design.
People are incented to do well, and if they don't do well, we don't do well. We use this for product feedback, as you're going to see. Actually, you've already seen. I'll reiterate. We find out who's a good reference and testimonial. NPS is a single question, right? How likely are you to recommend? Use the Likert scale, 0 to 10. A lot of opportunity to do badly, right? Very little opportunity to do extremely well, but it's a very effective gauge of true customer sentiment. Our NPS is actually what an NPS you would expect if you were doing B2C. In B2B, it's very rare that you see companies doing 15%, 20% or higher. Why? Because it's a work software. You don't get that excited. Our customers are very passionate. The distribution's fantastic.
Because so many people are so passionate, and there are people who are on the border, when you talk to the people who are actually on the border, you may call them up and you don't ask them why you got an 8. You go, "Hey, how you doing?" They'll say, "Hey, I just filled out your survey. I love you guys. I gave you an 8." You go, "Oh, thanks, man." Because obviously that's still great. Some people will never give you a 10. Some people think 8's perfect. The point is, we can focus on those people and make sure they're getting all the value that we want them to get and that they need to get. According to our partner, in my own experience, I've been doing NPS for years, we're at 53 in world class.
I think it's a fantastic place to start. When I saw this slide actually originally in recent presentation, I thought, "Great, my work here is done. They don't really need me." I think we can build and grow. Let's look at the actual distribution. The people we onboard in the last 12 months, vast majority are promoters, very excited, very passionate. Our opportunity is keeping that going. We have champion change, people move on. We're going to make sure that everyone's very passionate continually. Part of the survey is to ask our customers, and these are the decision makers, the Chief, the Commanders, or key influencers of the Program Managers, how satisfied are you with our ability to make an impact in your community, in helping you be safer, helping you locate gunshots more accurately, respond more quickly?
That's for the top three of these. Overwhelmingly satisfied or very satisfied. One of the things that I heard continually when I was on site, the thing that they will tell you, the officers, is ShotSpotter is a safety solution for them. Can you imagine if someone tells you, "Hey, there's gunshots. Go check it out." Are you serious? I'm not police, so not having accurate, timely information seems insane. People who've gone through the process of being skeptics, have implemented the best practices, and now have it integrated into their daily work life, of course they're going to use it and want it. They don't want to work without it. That's what the most satisfied customers agree, that it does give them the data to be situationally aware. Also, it helps with police and community relationships.
When I first joined, it's like, well, what does that mean, actually? Can you imagine you're in a neighborhood where there's gunshots, 80% of the times they're not even reported. Police don't show up in those cases. Now that you know about them, you can show up. Following our best practices, you can engage your community more effectively to get more evidence and to help them understand that you do care, and you're going to make progress. Do the time check real quick. I was kind of disappointed. I realized Sam had tremendous videos, and I've got text on a wall. This is the type of feedback that you get with people using the system, professionals, chief commanders. One of the things we tell them throughout the onboarding process and routinely is we care about your feedback.
We try to get the feedback, and then we use it to make the product better. You saw the ILS as an example, Investigative Lead Summary. Standard practice, but we're really good at that. Again, they will tell you how it's integrated into their life and how it helps them solve crimes, respond faster and solve crimes. A summary of this is 95% of the agencies responding agree or strongly disagree that we actually help them increase the number of shell casings, get evidence, identify shots, and then get the patrol to the right place at the right time. 75% strongly agreed that we improve their community safety, allow for better interviewing of witnesses. At every level, we have a great impact, and people are willing to provide us testimonials. This is more actually of the same, is other people.
Atlantic City telling us it helps us respond quicker, solve crimes. I'm not going to read all the slides to you. You have the access to the slides. One of the things My mom doesn't work for the company, but if she did, she would've given us that middle slide. What they will tell you continually is when integrated into their system, it makes them better at what they do, it makes them safer at what they do. It helps them solve crimes, get evidence. It's a fantastic situation for us. I'm going to wrap it up here. For customer success, what we want to have is a promoter, vocal champion.
When I started googling ShotSpotter and looking at the stuff that people say on the news about what we do, I was like, before I was getting like, people were saying, "Yeah, I could do my job 15 minutes faster." Now we're talking about having tremendous impact in the community. It's fantastic. We have people who are highly likely to renew. They're very passionate about the product. 95% say, the respondents, we definitely want to renew. Many want to be promoters or want to be advocates. They want to share their story. That's fantastic that we have people who are willing to step up, work with us, document those, and capture those.
Because of the expertise on the team, we're able to help the agencies that want to leverage more technology, to leverage that, to expand their program, to get more value out of the program, using NIBIN and CGIC strategies. Our team can help them do that and implement those best practices. Ultimately, I do think we'll be successful at helping more communities get more coverage and be safer. I'm looking forward to 2020. Thank you. Oh, thank you. Yeah. You got to beg for it.
Good afternoon, everybody. Do I need a mic? Yes? Okay. This is Jon Magin, I'm the VP of International Sales for LATAM. Today to be able to briefly talk about the significant progress we've made in Latin America in the last 18 months or so. Just a brief background on me. I started out my career living in Latin America, worked in business development and sales for almost 10 years. I have been many years in the public safety side on a global business development sales as well. When the opportunity came up for ShotSpotter, I researched the company, saw about their purpose, met the team. It just seemed like a perfect fit, I'm very glad I made that decision to come on board in April of 2018.
Before I go ahead, I'm going to give more of a brief overview, some of the macro drivers in the region. We're going to talk a little bit about the business development, how we've progressed and where we're focusing it on. More generalities, unfortunately, not going to be able to get into a lot of the detail specifics to customers and cities. I did want to talk more about some of the macro issues that are driving us and why we're in the market in Latin America. It didn't like that one. Brief overview, there are over 24 homicides per 100,000 in Latin America, and that's more than a five-fold, versus what the U.S. market is. Okay. That's across the region. 9% of the world population, they've got 40% of all the murders. Okay.
The need is definitely in the region. Crime is the number one public concern, actually replacing economic worries. That's putting a lot of political pressure on the leaders down there to do something about this crime wave and the violence. Crime costs 3% of the GDP in all Latin America. That's over $260 billion. I think we're about a little over 2% GDP when it comes to how much we spend on overall crime. We've made a lot of progress in the last 18 months. Basically, we blanketed the entire region, visited over 10 countries, multiple visits to more of the focus countries. We've hit all the different countries, stood back and said, "Okay, where do we need to focus our efforts?" We focused on these key four countries being Mexico, Colombia, Brazil, and Panama.
We've got a lot of places in Latin America where they have a lot higher gunfire rates, but they have no resources. Okay? These four countries are fairly major economies except for Panama, but it stands on its own with the canal. It's very economically viable. They've got the gunfire issues, but they also have resources to fund projects like the ShotSpotter. Interesting note, both Mexico and Brazil in the last year, they've come on board as presidents, and they both ran on law and order. Just wanted to point out those countries in the world with the highest gun rate, and Brazil, Mexico, and Colombia are in the top five. I'm going to go to the next one, which is a little bit more important on the region itself. Mexico has roughly 20 homicides per 100,000.
Panama's got a little over 10. Colombia jumps up to 24. Brazil is closer to 30. Brazil, just to put things in perspective, we had what? We were talking 13,000 homicides in the U.S. of a country of 330 million. Brazil's a little over 200 million. This past year, according to The Wall Street Journal, they had over 65,000 homicides, and most of that was gunfire related. Just to put things in perspective, all those countries in blue would add up to about 60,000 in homicides, and that just represents Brazil. Brazil is also between the eighth and 10th largest economy in the world. They do have availability to resources. Mexico is actually a fairly more concentrated area. Five of the states out of the 32 Mexican states represent 41.5% of all the gun violence.
That'd be Baja California, Chihuahua, Jalisco, Guanajuato, and the State of Mexico, which is around Mexico City. You've got other areas that are very tranquil, but the news hits, and especially a lot of the tourist areas, about the homicides happening. That's also from a political standpoint, driving, we need to do something about it. Also with Mexico, they spend millions of dollars on these command and control centers where they have all their police, and this is at a state level. They have their police, EMT, and fire all in one concentrated area, and everything comes in and dispatched to 911. It all comes into these C5, C4 communication centers. I think there's one state in Mexico out of 32 that doesn't have at least one.
Now, one of the things that we come in to hear, there's a lot of videos, a lot of license plate readers, a lot of software, a lot of mapping software, but not a lot of content. There's a lot of people staring at screens. We come in, and we can make things, the cameras, the LPRs point towards where the activity is when it comes to the gun violence. If they have already had the system, we can integrate into it. If they're planning to get it, we can integrate into it and give them content to look at and react to. I think we talked about this, how the socioeconomics, some of the poorest areas in the United States. This could be a map of anywhere in the region and the United States.
Most of the gunfire is happening in the poorer areas of the cities, in this case, Bogota, Cali, and Medellin, but it's very stratified towards the poorer areas of the city. Collaboration in the region. I said I've worked many years in some various different industries. I've never come into a situation where this, where I go to a trade show, start talking to competitors when it comes to, at least for the dollars, from software companies to hardware companies, security companies, and I think we already talked about NIBIN. This is one of the security companies. NIBIN doesn't work unless they can find the bullets, and we can help them do that, drive them to where the actual gunfire happened. When 80% goes unreported, that's a very important tool to have. They've opened up and say, "Hey, we'll introduce you to all of our customers.
We'll introduce you to our representatives in country. We're getting the same thing from some large software companies that do a lot of the mapping, as well as CAD type software. Hardware companies, cameras, LPRs, the same thing, where they're taking me in to have meetings with their customers because they want to give their customers more value for what equipment they have or they're planning to buy. We give a lot of content and help those cameras point and those LPRs point to where the actual gunfire is in a very timely fashion. Other places that we're dealing with and working with and collaborating, the U.S. Department of State, some of their departments, as well as the United States Department of Commerce, regional meeting services, Police organizations throughout the region.
We've recently done a major conference in Mexico, sponsored it. Also with NGOs, business councils, and some funding sources, some of the regional banks and larger banks to be able to fund some of our projects. Marketing. We've hosted on a couple of different occasions, multiple states security secretaries that come to the U.S. We've hosted them during our visits. These would be state security secretaries. We've also hosted in Mexico City, a reception for over 200 law enforcement officials. We participated this year in the Expo Seguridad with a booth. It's the largest exposition for security in Northern Latin America. Earlier this year, we also did the same in Brazil at the LAAD Defence & Security, and it's the largest security expo in all of South America.
The main thing that we're trying to do is anybody we talk to, they say, "Yeah," but they're all from Missouri. They feel like, "Show me." They want to be able to kick the tires themselves. What we're trying to do is develop in each one of those major countries of focus, a showcase, essentially a pilot, so they get using the product so the local cities in that country can go and make a visit. Today, what we're doing is we're hosting them in the U.S., coming up to areas in the United States and talking directly to our ShotSpotter customers. Oops. The last slide, and I'm going to talk verbally about proposals that we have out there.
Going into 2020, we have 10 proposals that are budgetary proposals throughout those four countries of focus, of which that represents a little over $5 million in annual recurring revenue. Of those, we're going to have one to two of those in 2020 come in in the first six to nine months. That's just to kind of give you a general on where we are with some of the financials in the region. We are focusing our efforts and planting a lot more seeds out there. With that, I'd like to bring Gary up here for the North America sales. Thank you.
Well, good afternoon. I've been asked to move briskly through this material. I'm Gary Bunyard. I head up our North American sales team. I have been in public safety technology most of my career. Had the opportunity in January of 2017 to join this great company and join this incredible leadership team that Ralph has put together. Like so many you heard from earlier today, it's been the most rewarding experience of my career, being a part of what we do in the communities we serve. What I'm going to do in the next 25 minutes-ish is I'm going to talk to you a little bit about 2019 and hopefully help you understand why we are so thrilled and happy with the finishing touches that we're putting on 2019.
Going to talk to you about the strongest, most powerful sales team in public safety and why we believe that. Going to talk to you a little bit about some controls and processes we're putting together, why we feel so good about those and the impact they're going to have on sales optimization. I'm going to turn our attention briefly to 2020 beyond, and help you feel as good, hopefully, as we feel about the future and the sales funnel that we have in place to support our growth in the years ahead. With that, let's look back at 2019 briefly. 2019 has definitely been a challenging year for us. We had several major procurements that were essentially canceled for reasons that were totally unrelated to ShotSpotter. We had one procurement we're going to talk about in a bit.
It actually took two RFPs to get to the point where they could make an award. Several major procurements that have either been put on hold or have slid to the right with respect to timeline. We also did not have the benefit, in 2019, of another major tier one contract, another Chicago, another New York. It's been a challenging year, but in spite of those challenges, we are very happy with where we are year to date and the steps we're taking to put the finishing touches on 2019. As of the end of November, we brought in 87 new flex miles. Those are roughly broken out evenly. Not roughly, broken out evenly between new customers and expansions. In addition to the new flex miles, closed several really important contracts for the company. You've already heard a little bit about one, Las Vegas.
We started a pilot several years ago in Las Vegas, 6 sq mi. Based upon the tremendous results that they've seen, based upon their leadership, a great deal of collaboration between the elected officials, law enforcement, and ShotSpotter, they made a decision earlier this year to greatly expand their commitment to ShotSpotter. That commitment came in the form of moving from a single year contract, 6 sq mi, to funding and moving forward with us with a three-year commitment. A three-year commitment that not only included the original 6 sq mi, but added 17.5 sq mi to that system. We now have a three-year contract, 23.5 sq mi that we're supporting in Las Vegas with great results for that community. Another real important procurement I'm sure you've heard a little bit about is Puerto Rico.
There's several things I wanted to share with you regarding this particular procurement. We rarely have the opportunity to go through a competitive procurement and validate the unique position that we have in the marketplace. Occasionally we do. Puerto Rico. As many of you are probably aware, service supported Puerto Rico for a number of years. We were in the process of looking at a renewal scenario before Hurricane Maria hit back in September 2017. We're all aware of the devastation that that island endured. It took a number of months after that hurricane, before they were able to really get serious about looking at gunshot detection again. As they did that, it became very clear, they made the decision they were going to have to go through a competitive procurement. They were seeking HUD funding. They determined that the requirements surrounding that funding scenario involved a competitive procurement.
They issued an RFP in July of 2018. After a very lengthy evaluation, they announced that they were canceling that procurement and going to put a new RFP on the street because they had not found any competition. There was no viable competitor other than ShotSpotter. Fast-forward to June of 2019. Issued a new RFP. They worked very hard to pull together interest in the procurement, weren't advertised. After the evaluation period, we received notice of award November 1st, letting us know that we were awarded the contract. Contract three years, just over 20 sq mi, covering San Juan, Bayamón, Trujillo Alto, and we were able to turn that award into a signed contract by the end of November. Excited about that, the real story behind that is they went out seeking competition. They could not find a viable competitor round one.
Round two, as we learned through the award notification process, they did have one viable competitor. We scored a 91 out of 100 in the technical evaluation. That competitor scored 72 out of 100 in the technical evaluation. Our proposal for three years was $4.2 million. The competitive proposal was just over $9 million for three years. Once again, we were reminded of the unique position we have in the marketplace. There was a company that cobbled together a solution. They struggled to be technically sound, to compete technically. They also obviously struggled to compete financially. The results of that procurement are not unique to Puerto Rico. In fact, three out of the last three competitive procurements that have resulted in a contract yielded very similar results with respect to the competition. Feel real good about that step in the marketplace.
Putting the finishing touches on 2019, we got six or eight opportunities we're still working. Some of these were in various stages of walking the contract through the organization to get signatures. We're not going to close all 30 mi represented by those six opportunities, but feel that we're positioned quite well to close 18 mi, maybe 20 of those miles by the end of the year. You take the 87 mi we've closed year to date, another 20 mi, and we're expecting to finish the year just over 100 sq mi. We feel real good about that based upon the challenges that we faced. When you look at that performance compared to prior years, going back to 2016, you see we are taking a bit of a dip in 2019 compared to 2018.
As I mentioned in the beginning, we did not have the benefit of one of those tier 1 Chicago-type, New York-type opportunities in 2019. When you go back and look at the prior years, and you back out Chicago and New York, you end up backing 70 mi out of 2018. You back 25 mi out of 2017. You back 46 mi out of 2016. When you do an apples-to-apples comparison without the benefit of one of those tier 1 contracts, we've got a very nice trend developing here and feel real good about what we've accomplished with the opportunities we had to push through that sales funnel. Sales organization. We've got a number of exciting things happening with the sales organization. Before I get too deep into those, I'll start out with just several objectives here, very quickly.
Focusing on expanding the breadth and the depth of our sales leadership team, looking at steps we can take to allow me to spend more time looking over the horizon. Definitely looking at creating a more scalable sales engine, if you will. Taking steps to allow the sales directors to focus more time doing what we want them doing. Last, but definitely not least, looking at steps we can take to push each product more proactively, more programmatically, especially as we add more products to our portfolio. I'd like to start by introducing you to two of the newest members of the team, starting with Phil Bailey in the back. Phil graduated from James Madison University, worked for the city of Baltimore as a cop, subsequently a sergeant for four years, joined the company in 2010.
I can tell you without any hesitation, the most successful, the most respected, the most disciplined, a true sales leader in this organization, and very proud to promote him into the new role over product sales. Also, I'd like to introduce you to Joe Rodriguez. Joe? Joe started out with the Miami Police Department. Joe worked for Tiburon a number of years ago, and at the time when Tiburon was truly the leader in CAD record systems in the marketplace. He subsequently led U.S. sales for Intergraph, a company you now probably may know of as Hexagon, and was with Deccan for a couple of years as VP of sales, and joined us in August, heading up our regional sales. I'm going to talk very briefly about what the steps to bring them on mean to our sales organization.
Real quickly, looking at the organization we've had in place, we've had essentially five regional sales directors reporting to me, along with a sales operations group, sales operations focusing on proposals, data analysis, demos, and what have you. With a couple of product specialists, they are passively available to support the regional sales directors. Key step we're taking with these additions is establishing a new role over the regional sales directors. Joe has stepped in that role, really helping the sales directors focus on their region, their territory, maximizing our ROI as they work that geography. Phil stepping in the role over product sales, taking steps to build a team to start driving proactively each product and the sales of those products following various sales programs through those regional organizations.
We're also adding additional members so that every product in our suite has an individual owner dedicated to driving the sales of that product through those regions. We're also looking at the regional sales directors and steps we can take to allow them to focus on what we want them focusing on. Part of that is, it's not really reflected here, is the large national strategic accounts opportunities. I'm going to take one of those. Phil's going to take one of those. Joe's going to take one of those. Each of the sales members of the leadership are going to own those national accounts. On the low end, we want to make sure, as we continue to get more segmented in our focus and the TAM that Sam shared with you, we're going to add some quota-carrying individuals that are focusing on the lower tier of the marketplace.
Those one or two square mile opportunities, more feet on the street, focusing on that lower section of the TAM. Supporting that, we took steps last year to bring on a new contracts manager, offloading that from me, from the regional sales directors. With Paul now on board, we're going to be taking steps to take more of the account management, move it over to customer success. We've already taken steps and are going to take more steps to take the whole renewal process and move that over to sales operations. Again, the focus there is allowing the regional sales directors to spend more time doing what we want them doing, and that's hunting, and that's hunting with a focus on our sweet spot. I've got a number of process control things I was going to share with you.
Based upon the time, I'm going to quickly just mention that we established a couple of years ago the concept of territory business plans. Each regional sales director has established a business plan for how they're going to manage their territory as a business. With Joe's leadership, he's already working with them, taking a much more granular view of each of those regions, looking at the TAMs, looking at each of the segments of the market by state, and going to be going through a very rigorous process of establishing territory business plans, building upon the work that has gone into this in the last couple of years. We made a big investment in Salesforce over the last couple of years. We're going to continue to expand that investment, turning to Salesforce to help us manage our business. Bottom line there is visibility, accountability, and analytics.
Really making decisions based upon the data and what the data tells us. As we build this new product sales organization, we're going to build a series of playbooks on how we're going to manage these products and drive these products proactively through these regions. Another key thing I wanted to talk briefly about, but I'm going to gloss over, is an opportunity management playbook and the wheels that are in motion there to look at how do we want to manage each opportunity. When you look at the steps that I'm going to quickly fast-forward through these. When you look at the number of gates we have to go through to manage an average opportunity, what can we be doing based on lessons learned to design the shortest path through those gates?
When can we say we're doing it right, we're doing it most effectively, and in some cases, bypassing some of those steps, but navigating our way through those gates as quickly and effectively as possible. Looking forward, two key pieces to this last slide. One is, what are some of the forces working against us, and what are some of the forces in the market that are working in our favor? I've just highlighted a couple of those here. What does the funnel say regarding our future? Looking very quickly. Budgets are always tight, but there has been a significant number of agencies existing as well as prospective, that seem to be struggling with budgets and budget pressures from the top. Decisions are being made above them that affect their budget and the amount of funding they have available for gunshot detection. What does that mean?
That means we've got to get more involved in helping them justify. We've got to help them understand the value proposition. We need to help them find alternative funding. Competitive distractions. We've talked a little bit about competition. What does that mean? That really means there's more companies that are posing as competitors. That doesn't mean we're losing to those companies. That means we have to spend more time helping that agency understand how to differentiate what we're offering from those other companies that say they're competitors. That means we may have to occasionally go through an RFP before that procurement department is comfortable with the decision that's been made. Department of Justice is a source of funding for a number of our customers. Under the current administration, they've made a ruling that they do not support sole source justification procurements. Everything has to be competitively bid.
That doesn't mean we're losing any deals. That just means we have to work harder and sometimes take more time to help these customers work through that process. On the positive side, Sam showed you some of it. We've got more positive results, more quantitative evidence of the efficacy of ShotSpotter than we've ever had today. Thanks in large part to our customer success organization, marketing organization, and sales. We're collecting a lot of really good data that helps these agencies get comfortable when they're considering this technology, and they need to build a business case to their administration. More regional tipping points. What do we mean by that? More and more examples where we're building a presence, starting with an anchor tenant in a given geography, and based upon the success that agency is having, more and more agencies want to come on board. It's happening today in Ohio.
A couple of years ago, we established Cincinnati as a great anchor customer in that state. Now we have Columbus, now we have Toledo, now we have Dayton. Dayton and others will follow. Florida is another example. Great success in Miami-Dade, Miami Gardens. You look over the last couple of years, Jacksonville, Fort Myers, West Palm Beach, Tampa, many other agencies following. Last example. I was in a meeting just a couple of weeks ago with a group of suburban law enforcement agencies outside of a large metropolitan customer. We're sitting down with six or seven agencies. We start out trying to help them understand the value of ShotSpotter. About 10 minutes in, they cut us off. They said, "You don't need to sell us on ShotSpotter. We know what you do, we know we have a problem, and we know we need you.
We're here together to figure out how you can help us get ShotSpotter technology. You don't have to sell us. Help us figure out how to get it. Seeing more and more of those examples in given geographies where based upon success, we're able to build upon the success that these agencies are having. Looking at our funnel real quickly. In this case, I'm just looking at Flex opportunities, the entire universe of identified opportunities in Salesforce that we are tracking. There's over 250 opportunities representing over 800 sq mi. Some of those go way out there. A very large portion of those is actually closable in 2020. It's not going to happen. A lot of those are going to slide to the right. As we do every year, we're putting the finishing touches on 2019, we're hearing offers 2020.
Joe's already working with the regional sales directors on what are our targets for 2020. They have identified from that pipeline about 60 opportunities that they feel represent the highest odd opportunities, positioned most favorably for 2020, representing almost 200 sq mi. Based upon that sales funnel, we feel really good about our position and our ability to set some goals of 130 sq mi-150 sq mi for 2020. Feel real good about what we've done in 2019. Really excited about the team that we're building, a lot of the processes and controls we're putting in place to optimize the sales performance, and based upon our funnel and the growing list of net promoters, advocates out there, and the momentum we're seeing in a number of geographies. Feel really good about 2020 and beyond. Alan?
Thanks, Gary. It's amazing what happens when you tell people there's no bonuses if we don't get back on schedule. I could probably just stop at this slide because I saw the scribbling in the notes that everybody was taking in terms of how things are looking and where we're going. I'm going to go over And some of you have seen a lot of these slides, so I'm going to probably go over them fairly quickly, at a lot of them, so we have sufficient time to have some question and answer period. Before I do, I do want to introduce two members of our executive team. They're non-speaking roles today. Jean Chase, our Corporate Controller. Go ahead and say hi.
Hi.
She's in the back. Hazel Velasco, our HR Director as well.
Hi.
They aren't speaking today, but they're absolutely critical and key for us in what we're doing and how we're building the company that we're building today. In terms of the business model highlights, most of you know already, we deal almost all of our products on an annual subscription basis, SaaS revenue. Almost 100% of our revenue is SaaS-based, with very high margin, low variable costs. We have leverage in every single category of our operating expenses, and I'll show you a summary of the finances at the end to demonstrate that. Very efficient sales and marketing. You saw a lot about what we're doing in sales and marketing, and there's a lot going on behind the scenes. When you compare us to other SaaS companies, where we only spend $0.30 per dollar of new annualized revenue, it's an amazingly efficient machine.
That gives us very low customer acquisition costs. Low customer attrition or churn, which leads to high revenue retention. I'm also going to show a new slide today that talks about cohort analysis in terms of where you can see when a customer starts, how those particular customers expand over the years. Lastly, we have very rapid cash payback of installation costs. Many of you know that I referred in the past, we generally have one-year contracts. However, because of our high retention rate, they renew year after year. However, if a customer did decide to not renew after one year, we would still make money on that particular project. We have a break-even point in less than a year. Our goal is to cultivate at least 10-year customer relationships.
We do have some customers who have been with us many more years than that, but that is certainly our goal, which gives us a very high lifetime value to customer. If we could keep an average customer at $400,000 a year for 15 years, that's $6 million. Incredibly high value to that customer. In terms of the staff's recurring revenue model, we have continued to have these quarterly revenues that have generally gone up and to the right. We do have some flatness. There is a lumpiness to our business. You can see that in a couple circumstances as well here, in terms of Q2 to Q3. This is what I was talking about in terms of customer growth by commencement year. For example, you can see that the customers that were at year-end 2015 went from $10 million to $25 million. What does that mean?
That is obviously affected by Chicago and New York. Those have continued, it also shows that not just in 2015 beyond, but if you take a look at the ones that started in 2016 or the ones that started in 2017, those customers don't turn off the platform, they generally add and they increase their footprints. It's something that allows us to have a very high revenue retention rate. Many of you have been with us and tracking us since we went public. When we went public, we talked about our CAGR growth rate, somewhere between 30%-35%. We said, "All right, well, where are we within that scorecard?" Granted, we had some challenges in 2019, but in terms of where we are from 2016 to where our guidance is for the mid-2020, the CAGR growth rate was 33%.
We feel pretty good about what we're doing and how we're getting there. It is lumpy. We talk about that a lot, but hopefully, you have a lot more confidence in terms of what we're doing, in terms of adding to the capabilities of our team. Not just for the sales and marketing, but in the R&D side, that's going to allow us to continue to differentiate ourselves and hopefully continue that upward revenue ramp as we continue to grow. It's not just revenue growth. We have had a fairly high revenue growth in terms of the 43% from 2015 to 2018. You can see that goes up there. Equally important is the gross margin percentages .
When we talk about a lot of our costs that are going through cost to goods sold being semi-fixed, things like our Incident Review Center, things like some of the telecom costs. While we continue to add revenue, that spread over those semi-fixed costs allows us to get more efficient from a gross margin perspective, and you can see the percentages are continuing to increase. If we just take points in time, on the far right there, from Q3 of 2018 to Q3 of 2019, we increased our gross margin percentages from 55% to 60%, or from $5 million to $5.9 million. That is something that we do see that we believe we can continue. Again, that will be somewhat lumpy when it goes quarter to quarter.
Overall, we'll continue to add revenue without adding an additional amount of cost of goods sold at the same rate that the revenue's increasing. We should see gross margins continue to improve. I apologize for the size of this, but I wanted to just highlight a couple of the key stats here in terms of operating efficiency. I'm going to focus your attention over to the right four columns. For sales and marketing, we spent about $2.4 million in 2018. In 2019, third quarter, we also spent about $2.4 million. You can see the additional resources that we've been adding. There is a bit of an effect in there that quarter to quarter, you get a little variation.
The percentages is going down as a percentage of revenue. Research and development went from 1.2% to approximately 1.3%. G&A actually went down 2.9% to 1.8%. Now, there was about $900,000 in one-time cost in Q3 of 2018, essentially, our costs have been flat. As we continue to grow revenue on the top line, we're being very efficient in how we're spending the money. From the sales and marketing, you heard Gary talk about a lot of the new resources and reallocation of the current resources that we have over the last year. If you were to compare that versus 2017 as a whole, where we spent $6 million for sales and marketing, we already started ramping up the sales and marketing in 2017 and 2018, where we spent $8.4 million.
A lot of those resources are now being allocated in a way that's going to allow us to drive top-line revenue. While we're driving top-line revenue with increased sales and marketing, all of the things that you heard Rob talk about and Paul talk about and the team, and talk about that are happening behind the scenes in our R&D group and technology group, we're still investing in those things. We're not standing still. Even though we're being efficient in the R&D OpEx lines, we have a seasoned team. We do continue to add to them periodically, we're not spending in a way that is unwise in the R&D side. Same thing from a G&A side perspective. Lastly, on this slide, I'd like to focus on the revenue retention rate down here, second row from the bottom. Almost 140%.
You could think about that as a same-store sales number. Net of any attrition for 2018, all the customers we had expanded their footprints by approximately 39%. I don't know very many companies that have a revenue retention rate that approaches that type of number. We already talked about the sales efficiency in $0.30 per dollar of new annualized contract value. I'm going to just talk, and we are basically reaffirming our 2020 revenue guidance. We went through this on our Q3 call, where we did a build-up to get to the $48 million-$50 million in revenue, where we're going to exit 2019 at approximately $43 million in our annual recurring revenue.
If you can think about that, we know that the actual revenue is going to be lower than that, but the annual recurring revenue, because a lot of the contracts that we add in the second half of the year are adding to our baseline for 2020. We start approximately there. From a domestic sales net of attrition, we're expecting to add $4 million-$5 million, less approximately $1 million in attrition. We do build that in every single year, somewhere between 2% to 3% of attrition. On top of that, we're talking about adding another $1.5 million in international revenue. Lastly, that top bucket in blue is Missions, security, other types of revenue as well. How achievable is this? We're already going into the year with the $43 million in annual recurring revenue.
You saw Gary's presentation about how good we feel about the pipeline. By the way, that is the first time since we've gone public that we ever shared details about the pipeline. Also, as he mentioned, we have good visibility, but things could get delayed when they actually go live, when we bring on miles, sell them from a bookings perspective to when Joe actually deploys them, can determine where we are in that range of the domestic revenue recognized that we get for 2020. Same thing with international. Jon referred to a lot of proposals that we have out there. If there's anything that we learned in 2019, it's that the international deals are taking longer than we expected. We feel that we've made appropriate adjustments to where our expectations are for 2020.
Lastly, in terms of missions and security, we talked about how we're seeing a large uptake in the missions now, and we hope to see that as well in security. In terms of our guidance for 2020, we feel confident that we're guiding appropriately to that $48 million-$50 million for top-line revenue growth. We don't guide to margins at all, but we do expect our margins in general to improve along the way as we continue to grow revenue. Lastly, before I turn it back over to Rob for a summary slide, then we're going to open it up to Q&A. Just a couple things real quick in terms of capital allocation. We get asked a lot about what is our strategy for mergers and acquisitions.
We often say we're not looking to do anything transformative, meaning multiple tens of millions of dollars that's going to change our focus. We are going to continue to look for things like we did with HunchLab. We acquired HunchLab, the technology that we've now turned into Missions. The total cost of that, including earn-outs, was less than $3 million. We believe that creates significant potential value in terms of the annual recurring revenue that that's going to bring in as we continue to expand those sales and grow towards that $100 million mark that Ralph alluded to. Lastly, although we don't necessarily like that we had the opportunity to do this, we did have an opportunity to do a share repurchase, and we did sell some shares earlier in 2019 where we netted over $11 million.
Circumstances were with our stock price that allowed us to do a repurchase of more than the shares that we sold for approximately $6.5 million. That is an update versus what we had given out in terms of what we had completed by the end of Q3. As of December 17th, we've now repurchased 260,000 shares at a total cost of approximately $6.5 million. With that, I will turn it back over to Ralph for a quick summary before we open it up for a question and answer period.
Great. Thank you very much, Alan. Just stay up here. Let's go ahead and flip to the next slide. Great. Really appreciate you all keeping us on time here so we have time for Q&A. Let's talk about it. In terms of investment themes we'd ask people to consider and take away after your time with us today is, the first thing is that we're targeting and going after a large and under-penetrated market opportunity. Sam shared with you that we're looking at over a $1 billion TAM across all of our product portfolios, and it's significantly under-penetrated. The good news for us is that, at least for the ShotSpotter Flex service, we have a significant first-mover advantage in a market with little to no direct competition. The gunshot detection space is a market that we effectively have created. We effectively own it.
There isn't any direct competition. You heard from our technology team, the significant technological barriers that we've put in place that protects that franchise, protects that market. Certainly, the experience curve benefits that we've been able to generate from Joe's shop experience in terms of how we apply that technology in the real world is significant. Our core database for both gunshot events as well as non-gunshot events that let us continually improve our machine algorithms is a very unique asset that really gives us a significant lead. I think lastly, from an NPS point of view, our brand reputation is a barrier to entry as well, especially as we start penetrating deeper into the law enforcement agency market where the early adopters are gone. We're now looking at more of the late majority of folks, and these folks are very conservative and risk-averse.
The idea that they're going to step off the ice and do something super radical is very rare. There will always be an exception or two. We look forward to seeing how those exceptions work out. If someone wants to go and try an unproven gunshot detection capability and actually deploy it, we're going to be looking with great interest to see how that actually pans out for them. I think hopefully with the deep dive we've done on technology, you can appreciate the fact that we feel pretty confident in our ability to deliver this capability in a very unique fashion. Probably the thing that Joanne asks me this frequently, she says, "Hey, what is the thing that people probably don't understand about your business?" From my point of view, we have this incredible vertical business model advantage.
The fact that we're able to focus in a very unique way, and we're able to grow revenues, and those revenues, just by the nature of what we do and how we take care of our customers, are extremely sticky, and that's backed up with our revenue retention. What's really interesting is our low customer discovery and creation costs. We have this advantage because we don't have to boil the ocean for customer discovery, right? We know who these 1,400 or so agencies that are out there that represent our TAM. We get to see them at IACP every single year. We don't have to have very expensive marketing programs, again, boiling the ocean for customer discovery. Once we discover a customer, converting that customer, we can do so in a very efficient way. I think our NPS and brand reputation is really important to that.
Again, that's backed up with our low sales and marketing costs per annualized revenue dollar shows that with $0.30 per dollar of revenue. Again, from a cost to serve point of view, once we capture that customer, our cost to serve is relatively low too, because the technology, as exciting and sophisticated as it is, it's already been invented. The beautiful thing about gunshot detection is there's not a big feature war that we have to add on to that capability. Effectively, what we do is very hard, but at the end of the day, what gets presented to a customer is very easy. It's a dot on a map with some metadata around it. You don't have to be in a feature race the way we would be if we were in an ERP business. That's really fascinating.
Lastly, of course, this is just an amazing place to work, and it's an amazing place to collaborate with customers and hopefully investors around getting at this issue of gun violence. All that stuff works together to make this, I think, a very interesting compelling investment opportunity. It certainly makes it a thrilling place to work and lean in on every day. With that, thank you very much for your patience and not having to take breaks so that we can get through it. I think we're actually 15 minutes ahead of schedule.
We said we have time for Q&A.
We've got a lot of time for Q&A.
We also have a tour of the Incident Review Center and a demo of the Mission software. For anyone that wants to see those after Q&A, we'll take you over to the office and split out into two groups and then swap it for anyone.
Yeah, 100%. We should probably split folks up so we can do it in parallel, right?
Yeah.
Yeah. The senior leadership team, maybe you could just come up here, and we can just be prepared to answer any questions that you have. I know you guys have been sitting on them for a while, so we appreciate your patience.
Up there. Maybe we're so thorough that we answered all your questions.
We're so thorough.
This is symbolic. The beach is so bright.
Yeah.
Yeah, right. Yeah. Maybe we can turn this off. Yeah. Great. Okay, awesome. Yeah, thank you very much. Yes, sir.
Jeremy Hamblin from Craig-Hallum. Wanted to just start with your guidance and think about this year, obviously, there was some slippage and discrepancy from where your initial guidance was good and a subsequent guide down from there. As you look at that very achievable range that you've set, the $48 million-$50 million, the various components of that, where do you see as the most at-risk? Would it be on the international side, the $1.5 million? Is it more the domestic miles, the $4 million-$5 million, net of $1 million attrition? Is it Missions and Labs? Where do you sense as the potential pinpoint?
Yeah. The question is in the guidance for 2020, where do we see the most risk in the elements? I'd certainly turn it over to Ralph too for additional comments, but I think we were fairly thoughtful and, in every single category, trying to be reasonable. We're not hypers. We don't try to sandbag. We try to be as realistic, but take the lessons learned that we have. I think out of all of them, we feel pretty good about the ranges that we've given from a domestic side, what Gary highlighted in terms of the actual pipeline and what he's targeting for his sales team. They have their comp structure tied to that as well. We feel pretty good about that. Timing is the biggest risk. In other words, we may have a fantastic sales or bookings, but we may have circumstances.
For example, we had three circumstances in 2019 where customer requests caused us to delay either the contract signing or the contract implementation, which adds up to hundreds of thousands, even more than $1 million in revenue. In terms of actual buckets of revenues, I would say it's more tied to time, not necessarily deal flow in terms of what we think we're going to get.
Yeah. I'd just say, if you want to just look at the domestic revenue segment, for example. If you were to theoretically light up 25 mi a quarter and just evenly over the year, so 100 mi at the end of the day, that would net $3.5 million in revenue. Maybe a little bit less because some of those deals are going to come in, not at the $70,000 per sq mi price, but at the $65,000. 100 mi, 25 sq mi a quarter, that $3.5 million. If the attrition is high, like $1 million, that could maybe put that at risk. If the miles are back-end loaded. I think we feel like we're going to be off to a fairly positive start in Q1 from a go live mile point of view. Again, as Alan said, we tried to be reasonable.
That's a big number, and it's definitely going to require a cadence of go live miles. If for any reason, Gary could do a great job on the book side, but if for some reason it takes a while for those miles to go live, it does have a GAAP revenue impact.
Just as a follow-up to that, then. You have probably decent visibility on Q1 at this point. Gary had a range that he mentioned of the target quota, not the potential, which was much larger, but the target quotas of 130 mi-150 mi. Presumably, if you hit the midpoint of that, you will easily exceed your guidance of $48 million-$50 million.
Well, again.
That's all if it's very much back-end loaded. Is that fair?
Yes. Absolutely. Yeah. That's the difference between the cadence of booked miles versus go live miles.
Thanks.
Yeah.
Yeah.
Your guys' medium-term goal is adding 600 gross domestic miles. How should we think about the activations over the medium term?
What we build into our expectations is an attrition of between 2.5%-3% every single year, right? Just round numbers, you would expect if we've got 700 mi, somewhere between 14 mi and 21 mi would be lost through attrition. It is our goal to not have any, but realistically, as a SaaS company, we know that's going to occur. When we build the guidance, we do build that into our expectation. Hopefully, we can maintain the percentage at 3% or below.
Two questions.
Could you talk a bit about going forward, what we should expect as far as the mix between expansions in new cities? I think Gary mentioned in 2019 it was even, like 50/50. I think that's been closer to how it was in the past. If we look at the medium-term targets, it's 5x the number from new cities versus the expansion. Could you kind of explain going forward that mix?
Sure. The question was from a mix of actual go-live miles near term versus what we see longer term on that march to $100 million. It's true. We saw more of, I would say, closer to 50/50 in terms of expansions versus new city go-lives probably in the last year, even year and a half. I would expect it's going to shift more towards what we see as that medium-term target, where it's maybe not 80/20 in terms of new cities. It's probably going to shift more like 60/40 this year and start adding more towards the new cities. That said, we do expect to have expansions continue, and it may be that ultimately in that 600 mi, and again, that's really just an aid for you to help think how we're going to get there.
It might end up being 200 mi and 400 mi or something different than that.
Yeah. Let me just comment on the 500 mi. The 500 mi includes initial miles as well as expansion miles for those new cities. It could be the case that technically expanded miles of the 600 mi might represent something on the order of 200 mi of the 600 mi are expansion miles, because those next 100 cities maybe initially come in at four square miles on average, and then there's 100 sq mi of expansion on the second 100 city cohort. Yeah.
Thanks. A couple of things. First, I heard a couple of times today that 2019 was a rough year, and there were some pushouts and stuff. We also heard a lot about adjustments to the sales organization and reallocating people's time. I'm wondering, to what extent was 2019 just a tough year, and to what extent do you think it could have been addressed if these sales organization changes had been in place? I'm just trying to get a sense as to, in particular, how we might expect your closing cycle to change or improve as a result of these changes to the organization.
Yeah. Maybe I'll take a crack at that question. I think at the end of the day, as Gary pointed out, we're going to have booked over 102 miles, which is pretty impressive. I think the challenge from a GAAP revenue point of view is from a go live point of view, a number of those miles don't count in our go live GAAP revenue mile numbers. I don't know that a change to the sales organization really would've changed the bookings number, and certainly the sales reorganization wouldn't necessarily affect the go live case. I can't see that, certainly for 2019. I don't know if you have a different point of view.
That being said, I think we're very constructive about the changes that we're making to the sales organization, along with the new investment that we're making in the customer success organization, which is really critical to, I think, our go-to-market capability. I don't think we should underestimate what Gary spent a little bit of time talking about, and maybe you can go into more detail. These regional tipping points that we see out there are extremely powerful. That's a real key consideration for us in terms of how we think about growing the business.
Absolutely. We see more of those developing today than we saw this time a year ago. It's hard for me to say, well, had we had this organization in place, how would things have been different? A number of the challenges we have, we've dealt with in 2019, were simply procurement issues, that it doesn't matter how we were organized, those things would have still unfolded. I can say with great confidence, we feel better about the pipeline we have going into 2020 than we felt this time last year in terms of the opportunities and the levers we have to pull and the number of opportunities we have to drive.
Thank you. Your answer sort of begs one follow-up, and then I'll shut up. I would posit that as a growth company, investors are probably as interested in your bookings and your miles as they are in your GAAP revenue, and in particular, what the future looks like. Alan, again, are we going to see some miles booked or ARR numbers from you that will give us a sense in sort of a point-in-time run rate?
Yeah. The question is, are we going to start reporting on bookings? I think the short answer is no. We are considering ways to add more metrics, though, that are meaningful to you, everybody, every investor analyst that we have. We have started to introduce those, such as the ARR and the first introduction of bookings. I don't want to commit to that at this point, but certainly, we're looking at ways to tell this story in an effective way that people can see the energy and the motivation we have in a way that is maybe more classically seen by you guys.
Thank you.
Back around the source again.
Now that you have more proposals out there on the international side and more experience in those target forward countries, I'm just curious if you could describe the differences between the international and the domestic sales cycle or any challenges that you're seeing now that you have more experience there?
Well, I'll go ahead and start, and then I'll hand it over to Jon. I would say, they're definitely longer. We know that. We had originally expected the international deals based on some early indications we had, might have been six months-ish longer than our domestic sales cycles. I think it's clear to us now as we're continuing to go after some of those that we could have been off by a factor of even 100%. It might take two years, it might take two and a half years to do some of those initial city sales in new countries. We are hopeful, though, and I'm going to turn it over to Jon at this point, that once we get some of those testimonial type cities in these new countries, that that can start going from there, and maybe shorten the sales cycle.
Any other thoughts there, Jon?
Yeah. International, especially Latin America, it really has been, I've got a sales title, but it's a business development. Some of these people have never heard of our system before. Just to get them to the point where they actually want a proposal and to make it happen, it's taking a lot longer. We did initially have some customers that jumped out of the woods very early in my tenure here at the company, as it's been less than two years.
They were giving us all the signals that they were going to go, and then they had budget cuts, and it didn't materialize. We thought, boy, if that's the indication, it's going to be fairly similar to the U.S. I've been doing international for most of my career, and it usually is a longer sales cycle, and especially in the Latin America areas.
I have kind of an opposite question to what was over here, which is, you guys mentioned, if I characterize it, as procurement issues during, at least domestically, during 2019 that made it a more challenging year and not necessarily things that internally was poor execution. Broadly from that standpoint of, why make change and why add a bunch of additional team members on customer success and on sales and these things that traditionally you have done really well in just a challenging year. Why does it make sense to sort of restructure sales and add sales and add folks on customer success when those are some of the biggest strengths for the company?
I'll walk through some of those organizational changes pretty quickly. What is really behind that is building a more scalable sales engine, if you will. We're not necessarily putting more direct sales reps out in the field, but I like to refer to it as putting more wood behind that arrow, making each of them more effective. We are also supplementing them, as I mentioned, with some lower-tier, sales-focused, additional feet on the street quota-carrying people that are out there pursuing some of the lower-tier opportunities so that the sales directors that are in place today can focus on our sweet spot, those opportunities five, six, eight square miles that have been our bread and butter over the years.
It is my expectation, putting more wood behind the arrow, having the organization that Phil is stepping into, but is going to be building here this year, is really pushing these products programmatically. Each individual product with owners is going to allow us to do more to push those products than would otherwise be the case, and I think is going to allow us to accelerate opportunities in that sales funnel long term. Is it going to make a difference in Q1? No. It's going to take some time. I do think it's going to allow us to basically expand the capacity of that sales engine and drive the opportunities through that funnel faster.
Yeah. Is it fair to say that the quotas are going to be very similar in 2020 as they were in 2019 from a miles point of view?
Yes, sir. I hope so.
Yeah. What were the miles quota for 2019 was?
150 mi allocated across all of the territories.
It's not like we're ramping that up significantly, even with the addition of these additional resources. It's about making the existing territory sales reps more capable of being able to hit that 130 sq mi-150 sq mi. In terms of customer success, we've always talked about the idea that that was going to be an area that we were going to invest in significantly. As we add customers, it's really important that we resource our customer success organization to be able to drive net promoters out of those customers that we're adding on, and we're adding customers at a fairly healthy rate. It only stands to reason that that organization is going to grow. Certainly, having some focused leadership around that organization was critically important that can bring that NPS customer success philosophy into the leadership team.
That was something new for us to do, too. Originally, that customer success organization reported in through sales, and now we've elevated it to the senior leadership team. We thought that was strategically important.
Yes, sir.
Up the road in Piedmont, the city just spent three years and over half a million dollars arguing with Crown Castle just to add 15 micro cell densification. The second time through, with a shot clock from the FCC, the thing happened really quickly. In your permissions business, do you have any leverage other than just a commercial negotiation?
That's a great question. It's something we're actually trying to work towards all the time. Some of our bigger partners that we've used multiple times in some geographies are the people that own and operate the lighting infrastructure, the power infrastructure, utilities, and whatnot. When we've got a relationship with them, which often can take quite a long time to get through their bureaucracy to get the approvals necessary to be able to put our sensors on their plant. The repeat times to go back and get those, the second city, the third city, the fourth city, especially if you're looking at these little regional tipping points that Gary's talking about, becomes really, really advantageous. That's probably the biggest area that we have been pursuing and continue to pursue.
If we find other opportunities, where you've got major property holders, we've done a couple things in some of our larger markets with major property management groups, real estate investment firms, that may be another field of opportunity for us. It's something we definitely look at.
I assume you don't run into the public's health concerns about RF emissions and all those things.
Oh, yeah. No, we do not.
Probably have time for one more there. Sure.
You guys highlighted how important Chicago and New York have been over the last four years. Do you think it's possible you could win another city on that scale in the medium term, or is there something unique about the population density or the topography of those cities that's?
Different from Los Angeles, Houston, Dallas.
Yeah. I think Alan and I, when we've been on one-on-ones and road shows, we've always talked around the fact that we should not expect another 70 sq mi or 115 sq mi customer that's representative of NYC and Chicago. I think to the extent that we add additional kind of tier 1 cities, if we can pop those things up to 25 sq mi, 30 sq mi, that's a big win for us. I think from our own internal planning purposes, we're not thinking we're going to see a New York City or Chicago scale.
What's the difference? Because Los Angeles and Houston are certainly as large as Chicago are.
True. I think in the case of Chicago is its own thing in terms of the level intensity of gun violence. I think in many ways, New York City is its own thing, too, in terms of its level of sophistication in leveraging technology and ability to have resources to invest in something like ShotSpotter. Is it possible that a L.A. or Philadelphia could be maybe 50 sq mi or something? It's possible. Not certainly in my thinking of around that. I think, again, if we got those cities at kind of 30 sq mi, 35 sq mi , that would be a huge win for us. We don't want to get over the tips of our skis in terms of expecting big numbers from tier 1 cities, because I don't think it's realistic or practical, quite honestly.
Yes, sir.
Just as a follow-up to that point, Philadelphia is an interesting use case. That's where I'm coming from.
If you look at the statistical evidence, they're seeing homicide rates have increased like six or seven consecutive years. They're using one of your competitor's systems that's kind of been grown a little bit homegrown.
It clearly is not working, and I can tell you there's widespread frustration in the community.
We agree.
How is that conversation now changing for the better? There hasn't been a change in terms of the person who's running that particular piece of the police and community.
Maybe that's the hurdle right there. Is that conversation, when you look at the statistical evidence that New York and Chicago, where homicide rates have fallen dramatically over the last three or four years, is that conversation getting easier to have, or is there no change because you don't have an internal advocate yet?
Yeah. I think it wouldn't be wise to talk about any specific engagement or city, so I'm going to have to demur from that. I think your question's a very good one in terms of how do we leverage the successes of Chicago and New York City in conversations with other agencies. Those are really important. I think particularly Chicago and New York City are quite relevant to those kind of tier 1 cities, those big cities that can more identify with Chicago and New York City, because a lot of our kind of tier 2, tier 3 cities, you got to be very careful not to put up Chicago, New York City, because it's like, okay, that's those guys. That's why the Cincinnatis are so important to us, the Sacramento cities are so important for us.
Medium-sized cities that are knocking the cover off the ball by leveraging ShotSpotter and other technologies and processes to prevent and reduce gun crime. We're not giving up on Los Angeles, we're not giving up on Philadelphia. We're not giving up on any tier 1 city you want to point to in the state of Texas. That's a huge opportunity. We don't have one deployment in Texas. That's a big problem for me. We have to be in Texas and be relevant there, and we know once we get there with that first anchor tenant, good things are going to happen for us in Texas, because there's a number of cities there that have a gun violence issue. We're going to keep kind of plugging along. You see this team here. You see the technology. You see the pipeline.
We've got a great road in front of us that we control. We're just going to be out every single day, just kind of executing and growing the business in a responsible, profitable way. I think that's all the time we have for questions, but I think I would appreciate it if anyone would like offering feedback. Is this a helpful time for you all? Any feedback?
Absolutely
on this? Okay. We're good? Okay, great. We're going to walk over then and do for those of you that have the time, we can walk over to our offices and do the Missions demo in the IRC if you'd like.