Good morning, everyone. Welcome to Cognex's 2025 Investor Day. We're thrilled to have you with us, whether you're joining us here in Natick or tuning in via webcast. My name is Greer Aviv, and I'm the new Head of Investor Relations here at Cognex. Just for housekeeping purposes, the full presentation is available on the Investor Relations website, and a recording will be posted after the event. If everyone in the room could just take a minute to please mute your phones. A quick note that today's presentation will contain forward-looking statements regarding our expectations, which are subject to risks and uncertainties. Please refer to this slide for our safe harbor language regarding forward-looking statements. Today is all about giving you a deeper look into the strategy, innovation, and people that power Cognex and set us apart in the market.
To start the day, Rob Willett will take the stage to present "We Are Cognex," with a focus on who we are, our culture, and our 2024 accomplishments. Then Matt Moschner will share our market opportunity, his vision for the future of Cognex, and outline our long-term strategic objectives and preview the financial framework that supports it as he steps into the role of CEO. From there, Reto Wyss will walk us through the cutting-edge AI technology that drives our innovation. Shirin Saleem will follow with an overview of our comprehensive ecosystem, and Carl Gerst will explain how our direct sales model gives us a competitive edge and drives customer experience. After a morning Q&A session with Reto, Shirin, and Carl, we'll break for product demos and lunch. Please note that this portion of the event will not be webcast. The webcast will resume at 1:15 P.M.
When Dennis Fehr will present the financial framework in detail and the capital allocation strategy that supports our growth strategy. We'll wrap up the day with an afternoon Q&A session with Rob, Matt, and Dennis, giving you the opportunity to ask additional questions. Thank you again for joining us. We're excited to share our journey with you and to show you how Cognex will continue to lead the way in AI-powered machine vision technology. Before I welcome Rob Willett, CEO, to the stage, please take a moment to enjoy this brief video about Cognex. Thank you.
Advanced, but not easy. Easy, but not advanced. To be truly transformational, technology must be both. When it comes to machine vision, only Cognex delivers both. By combining advanced technology with exceptional customer experiences, technology backed by decades of research and development, with about 1,400 patents issued and pending, and a track record of pioneering AI-driven vision systems since 2017, we have also done the hard work to make machine vision easy, with single-button setup, guided workflows, a comprehensive ecosystem, and expert global support. With proven AI tools for identifying, inspecting, gauging, and guiding, Cognex products are trusted in the manufacturing of 2 billion products a day, helping to ensure quality, lower costs, and improve results. That trust is built on unmatched reliability and intuitive operation. To keep businesses moving forward, we continue to redefine limits and accelerate productivity, all at the speed of easy. Cognex, advanced machine vision made easy.
Good morning, everyone. Thank you for joining us. Today, you will hear from Cognoids about progress we are making in our business and with our technology. I'll begin with key messages which you will hear throughout today's presentations. As a high-technology growth company, Cognex is built on a foundation of innovation. Our long-term commitment to R&D is a core strength that sets us apart. We're proud to be recognized as the technology leader in our industry. Our leading-edge software and AI tools, the accuracy and speed of our vision systems, our deep domain expertise, and the intense, close technical relationships we have with the world's leading manufacturers of discrete products give us a strong competitive advantage and create significant barriers to entry.
Today, you'll hear how we're focused on leading in AI-enabled machine vision technology, providing the best customer experience and executing on plans to double our customer base over the next five years. You will also hear how our financial strategy will focus on high-margin growth, strong cash flow generation, and a disciplined capital allocation approach that supports long-term value creation. Those of you who are here in person will experience some of our unique motivational culture. As Peter Drucker is credited with saying, "Culture eats strategy for breakfast." Our culture allows us to attract, empower, and retain the best talent in an industry where innovation and intellectual property are key to success. Many of you are already well acquainted with our company, but for those of you who are just getting to know us, let me provide a brief introduction. Cognex is a technology and growth company.
We have an exceptional brand based on our leading technology, deep customer relationships, and a direct sales approach with more than 30,000 customers. In 2024, we generated more than $900 million in revenue. We have a global footprint with about 2,900 Cognoids around the world and a revenue mix that reflects our global presence. Over the past decade, Cognex has consistently delivered strong financial performance, achieving an average adjusted EBITDA margin of 28%. You'll hear more from Matt and Dennis about how we're sharpening our focus on bottom-line profitability and how this emphasis fits into our broader long-term financial framework. Turning to what we do, everything we do is based on machine vision, the technology that gives computers and automation equipment the ability to see. A vision system operates much like human vision. Your eye is a sophisticated optical device, like a camera.
It captures information and sends it to your brain, which makes sense of it. In machine vision, an imager, lights, and lenses are used to capture data. That data is sent to a processor where vision tools or algorithms interpret it. Software is at the core of Cognex technology. It is the brain of machine vision, and we are a company of cognition experts, from which our name, Cognex, derives. Replicating vision with the 99.99% accuracy required in manufacturing takes a lot of technology and know-how, which Cognex has spent over 40 years developing. At Cognex, we focus on machine vision for the most challenging applications performed on high-speed production lines in manufacturing and logistics. It's something we do better than anybody else. It is this technology leadership and culture of innovation that have led to a strong track record of growth.
It has resulted in nearly 15% annual growth over most multi-year periods throughout my tenure as CEO. As you can see, our business has a cyclical nature, often influenced by major technology transitions, and growth has therefore not always been in a straight line. Due to COVID-era overinvestment, combined with high interest rates and a more challenging macroeconomic environment in some of our end markets, the past few years have presented headwinds for Cognex and our competitors. That said, we returned to growth in 2024, driven largely by inorganic expansion. Later today, Dennis will walk you through our growth formula, including the expected contributions from each of our vertical markets. Let's stay with 2024 for a moment. It was a year marked by strong execution on our three strategic priorities.
From advancing our technology to expanding our customer base and strengthening operational discipline, 2024 was a pivotal step forward, specifically on those three priorities. One, in 2024, we expanded our portfolio of machine vision products powered by world-class AI, such as the In-Sight L38, the industry's first AI-enabled 3D smart camera. Our newest products not only provide powerful functionality that allows customers to solve problems in a more human-like way, but also are significantly easier to use, integrate, operate, and sell. These products are, as our slogan says, "Advanced machine vision made easy." Because they are easy to use and sell, they allow us to reach less technically proficient customers who have historically been unprofitable for us to serve.
Two, we are evolving our sales force to bring our newest technology to a significantly larger universe of customers, one that we estimate is 5-10 times greater than the 30,000 we serve today. This new structure has already delivered strong results. The original class of entry-level sales noids that we trained and deployed to the field in the start of 2024 completed over 80,000 customer visits, ramped to sell $1 million per week while also referring nearly $10 million in business to our other sales engineers. This program is significantly expanding our market coverage and unlocking meaningful growth opportunities. The cohort we hired in mid-2023 added more than 3,000 new customers in 2024. These early outcomes validate our strategy and demonstrate the power of aligning talent, technology, and structure to drive scalable growth.
Building on what we learned, the second class of these sales noids were hired in mid-2024 and entered the field at the start of the year. They are ramping nicely. Three, last year, we successfully integrated MORITEX, the largest acquisition in Cognex's history. This milestone enhances our ability to deliver more complete machine vision solutions, combining advanced lighting and optics with our industry-leading systems. It has significantly grown our presence in Japan and the scale and competitiveness of our semiconductor business. Importantly, the integration has been smooth and is already contributing positively to our bottom line, reinforcing the strategic and financial value of the acquisition. I am pleased to introduce Matt Moschner. As many of you heard on our Q1 earnings call, Matt will succeed me as CEO effective June 27th, 2025.
This transition is the result of a multi-year succession planning process in close collaboration with our board of directors. Since joining Cognex in 2017, Matt has quickly risen through our leadership ranks. He was one of several high-potential leaders developed through our succession and leadership development program. Over the years, Matt has successfully navigated both challenges and opportunities, demonstrating strategic vision, operational excellence, and a deep understanding of Cognex and our markets. The board and I are confident that he is the right leader to guide Cognex into its next phase of growth. Without further ado, please join me in welcoming Matt Moschner.
Thank you, Rob.
I want to begin by first thanking Rob on a personal note for his mentorship over the last many years and the incredible legacy that he leaves behind at Cognex. Thank you, Rob. I also want to thank all of you, our investors, our shareholders, our partners, and other stakeholders for being here today, for your continued support and interest in Cognex. The last few years, as Rob mentioned, have been challenging, but I'm excited to step into the role of CEO at what I feel like is an incredibly pivotal time for the company. Since I joined Cognex in 2017, I've seen firsthand the power of our technology, the strength of our culture, and the passion our teams bring to solving some of the most complex manufacturing and automation challenges in the world. Looking forward, I only see potential.
Cognex has always sat at the intersection of breakthrough technology and real-world tangible impact. It's what I find so exciting about working at Cognex. Today's market environment feels different. Technology is changing faster because of AI. The world's supply chains are reshaping in ways that maybe we wouldn't have even imagined 12 months ago. Our customers' needs for automation and machine vision are deepening because of rising labor costs and regulations, and our competitive dynamic within machine vision continues to evolve. It's times like this that Cognex and Cognoids are, frankly, at our best. Today's agenda is designed for you to see that potential up close. I'm convinced that when you do, you will feel like I feel every day, which is immense excitement for where we are as a business today and where we can get to in the future. Let's get started.
What's the most basic question I think about every day? What makes us so unique? There are many things. The first and foremost, we are the recognized technology leader in our industry. Our reputation is built on decades of domain expertise, and our brand, Cognex, is synonymous with excellence in machine vision. We back this up with a committed investment in R&D, especially in the areas of embedded computing, optics, and, of course, advanced AI. For us, it's not just about keeping up with our industry. Our job is to actively shape our industry. Second, we operate within a large, growing $7 billion market, one that spans a diverse set of industry verticals, customer types, geographies, from logistics to pharmaceuticals and automotive to consumer electronics.
With a projected 10-11% market CAGR through cycle, it gives us a strong base to build from in reaching our own growth ambitions. Third, our direct sales model allows us to connect with end users at every stage of their buying journey. This hands-on consultative sales approach enables us to more deeply understand customer needs and build sticky long-term relationships that drive loyalty and repeat business. For those that joined us at yesterday's pre-dinner, you would have gotten a sense from the customer panel that we had. Fourth, we work with some of the world's most sophisticated machine builders and end users. These partnerships, many of which span decades, are a testament to our credibility and technical excellence.
As you will hear later today, we see significant potential to expand our customer base even further, as much as two times larger than it is today over the next five years. Fifth, our software is embedded directly on device, and those devices are tightly integrated into our customers' operations. The combination of these two things allows us to command software-like margins, which is a powerful driver for our own long-term profitability. Sixth, our capital-like business model enables us to generate consistent, high-quality cash flow while maintaining a robust and healthy balance sheet. This financial strength gives us the flexibility to invest in growth, return capital to shareholders, and navigate uncertainty without compromise. Now, Rob mentioned it. Underpinning all of this is our unique culture. It drives how we collaborate, how we solve problems, and how we continue to push the boundaries in an evolving industry.
Let's double-click into the market verticals that we serve. At Cognex, our success has been driven on a laser focus on the right industrial sectors, those sectors where our technology delivers the most value and where demand for automation continues to accelerate. Our journey as a company began in semiconductor, where precision and reliability are paramount. We expanded next into automotive and electronics, where automation drives quality and efficiency. More recently, we've made significant inroads in logistics, a sector undergoing rapid transformation through e-commerce and other supply chain innovation. Later today, Dennis will walk through our growth expectations by each of these market verticals. He'll break down how much of that market growth is driven by underlying market dynamics and how much is expected to come from increased vision penetration. We'll unpack what that means. Let's start with logistics, which is now our largest and fastest-growing end market.
After a brief post-pandemic pause, where the industry took time to absorb a lot of the excess capacity that Rob mentioned, growth in this sector has sharply re-accelerated. In our Q1 results, we reported double-digit year-over-year growth in logistics revenue. That marks our fifth consecutive quarter of growth and the highest level we have seen since Q1 of 2022. The broader trend towards online order fulfillment is clearly here to stay. From an automation standpoint, we believe logistics is still very much in the early innings. You will hear more about our strategy to capture this growth during the logistics product demos later today. Let's move to automotive, which has been historically one of our largest and most enduring verticals over the last several decades.
Now, machine vision in automotive is deeply embedded in almost every stage of vehicle manufacturing, from measuring inbound parts to guiding robotic assembly to inspecting fine details like the leather stitching on a new seat. This market, as we've talked about before, has faced significant headwinds last year, and we've seen that softness continue into 2025. We expect a more moderate decline this year, while we remain very optimistic about the long-term outlook for automotive. Why? The increasing complexity of vehicles, higher inspection requirements, and a transition to electric and autonomous vehicles all play to our strengths. You'll see this firsthand during our automotive product demo session later this afternoon. Next is packaging, which has emerged as our third largest market vertical. This segment, just as a reminder, includes both fast-moving consumer goods as well as healthcare-related industries.
Growth in packaging has been driven by increasingly stringent regulation, traceability, quality, and compliance requirements, making machine vision not just valuable, but essential. We also see significant growth opportunities in this market through additional penetration. It is why we have been investing to transform and expand our direct sales force to reach a broader section of customers in packaging with products that are easier to use and faster to deploy. You will hear more about our market creation opportunities from Carl and during the packaging demo session over lunch. Let's move on to consumer electronics. While we have not seen major form factor changes in recent years, we believe the next wave of innovation in this market is on the horizon. The emergence of wearables, foldable devices, and other AI-centric consumer hardware could drive outsized growth.
These next-generation devices will be produced at massive scale and will be increasingly complex to manufacture, making precision and reliability more critical than ever. At Cognex, we have long-standing relationships with some of the largest and most sophisticated players in consumer electronics. As they bring new technologies to market, we expect to remain their partner of choice. That brings me to semiconductor, an industry where Cognex has played a critical role for decades. Our acquisition of MORITEX in 2023 brought with it even more depth with key semi OEMs. Together, our advanced machine vision technology is essential to modern chip manufacturing. We continue to see strong, widespread growth in semi, fueled by demand for high-bandwidth memory chips, which are foundational for next-generation computing and AI infrastructure, which is being built out as we speak.
With our proven technology and deep relationships, we think we're well-positioned across all of these verticals for growth into the future. Let's turn to our served market. Just as a reminder, we focus at Cognex on our served market, which we define as the total value of business that our products are designed to win. In other words, it's not just the overall market size for machine vision; it's the portion of that market where our products are directly applicable. Today, as I previously mentioned, we size our served market at approximately $7 billion globally, with a balanced split across market verticals that I just described. Cognex holds a strong competitive position in this overall market, and that share position we think is in the mid-teens.
On the right-hand side of this slide, you can see the key developments which have led us to revise this estimate since the last time we updated you on analyst day 2022. We have added new market segments, we have modified some of the vertical groupings, and we have adjusted for exceptional events like COVID. Now, there are several key secular growth trends that drive machine vision adoption across each of these verticals. Let's talk about each of them. First, manufacturers are under constant pressure to do more with less. Machine vision delivers the precision and consistency needed to reduce waste, minimize defects, and drive down costs without compromising quality, frankly, on products which are increasingly difficult to make. Smaller components and tighter tolerances place an added emphasis on automation in production. Second, thanks to our investments in AI and other technology, advanced machine vision is easier to deploy than ever.
What once required complex integration, specialized expertise, is now much more accessible and approachable by our customers. This means faster returns on investment, lower project risk, which allows them to think much more ambitiously about where they apply vision. Third, a shrinking labor pool and aging workforce are reshaping global manufacturing. According to a recent Deloitte study, the U.S. manufacturing sector alone is expected to face a shortfall of labor of 2.1 million unfilled jobs by the end of this decade. As a result, the cost of labor has been increasing at a faster rate than productivity. This gap, this growing gap, underscores the need for automation, and it is a powerful tailwind for machine vision growth. Fourth, as companies around the world evaluate more regional and potentially less global supply chains, automation becomes table stakes.
Think of this trend as almost an amplifier of the other three, adding only more urgency to customers' needs to adopt vision across their operations. In this market context, I'm moving intently to evolve our value proposition. Over the next few years, I will transition Cognex from a traditional provider of machine vision products to a future-ready, AI-driven vision platform which can serve more customers and deliver the best customer experience in the industry. For decades, Cognex has won because of three things. We have the best vision tools to guide, inspect, gauge, and identify with unmatched speed and precision. We have financial stability to give customers the confidence that when they work with us, we can support them in the near and over the long term. We are a global business, and we have a global presence to enable them to standardize their operations across continents.
To be clear, these things are still core to our value proposition and will remain so. As I mentioned, the world is changing, and so must we. There are three areas I will be most focused on. The first, we're taking our nearly 10 years of advanced AI leadership to the next level. From creating new, game-changing vision tools to reimagining the product design more holistically to streamlining how customers get support, advanced AI will broadly reshape how customers experience Cognex. A good example is human visual inspection, something we've talked about over the years. Today, we think there are tens of millions of people deployed in manufacturing facilities around visually inspecting products made. In the past, with traditional vision tools, replacing these human operators has been very challenging given the sheer diversity of defects that they aim to avoid.
With today's AI tools, we see a path to addressing these applications and unlocking new applications for Cognex vision. Next, we're going to make machine vision easier to use and easier to deploy than ever before, delivering a more seamless customer experience from pilot to production. This ease of deployment not only enhances the customer ROI, but also lowers the barrier to adoption across a broader cross-section of customers. A good example of how we're already doing this is a product we recently released, the DataMan 290 series, which you will hear more about later today. With this product, we have completely reimagined the setup experience, designing a highly intuitive workflow and the ability to auto-configure the system with a single button press. What has historically taken hours now takes minutes. With this intense focus on ease of use, I will transform the business.
Finally, we're launching a comprehensive product ecosystem with hardware, software, services, and support designed to help customers standardize on Cognex globally. This actually is a strategy that I have been leading for several years while leading our product and engineering teams. Our ability to develop technology in a common way not only helps our customers move more easily between our products, but drives stickiness with them, all while letting us be much more efficient with our own R&D spend. This evolution of our right to win will let Cognex redefine what's possible in machine vision and continue to build the future of automation. At the same time, we will transform our go-to-market and broaden our customer base through targeted investments in our sales force. As we've mentioned before, we think of our market as a pyramid.
Historically, we've done very well with customers at the top of that pyramid who are some of the world's most tech-savvy and bring some of the most complex manufacturing challenges. They value our technology, and they value our highly collaborative working style. Today, as Rob mentioned, we serve approximately 30,000 customers and believe that below that line lies as many as 5-10 times more who are smaller, more regional automation players. Through our investments in our sales channel that we started in 2023, coupled with innovations like edge learning, which make vision easier to deploy, we've added approximately 3,000 new customers across geographies and verticals. We're seeing promising early wins from these investments. Over the next five years, as we've mentioned, our goal is to double the number of customers that we serve.
I have three strategic objectives for Cognex over the next five years, which I already mentioned during our Q1 earnings call. First, we will be the number one provider of AI technology for industrial vision applications. We will accelerate innovation by continuous investment in AI, product development, and platform capabilities to stay ahead of customer needs and lead industry trends. Second, we will provide the best customer experience in the industry, delivering seamless engagements from first interactions to full-scale deployments through our direct sales model, a unified product ecosystem, and significantly upgraded customer support capabilities. Third, just to repeat myself again, we will double the number of customers we serve by scaling our go-to-market engine and reaching new segments, geographies, and verticals.
Now, if you take these three objectives together, I expect this focus to yield significant results, including us achieving a number one or number two position in all major markets that we serve. We have designed today's agenda to clearly outline and support this strategic direction. Each speaker will highlight a key pillar of this growth strategy. Starting with Reto Wyss, we will showcase our innovation and leadership in AI technology, emphasizing how it drives differentiation and long-term value for customers in Cognex. Shirin Saleem will explore the power of the Cognex ecosystem, illustrating how our integrated solutions deliver measurable benefits to our customers. Carl Gerst will explain how our direct sales model creates a competitive advantage, enabling us to expand our reach and attract new customers. Now, the successful execution of these objectives will fuel sustained growth and ensure expanded profitability, which are both critically important to us.
At a high level, our financial framework is designed to deliver consistent long-term value creation through economic cycles. We focus on three core metrics that reflect our disciplined growth, operational efficiency, and capital allocation strategy. First, we target a 13-14% compound annual growth rate in revenue, driven by innovations and strategic market penetration, and an approximate 3% contribution from M&A. Second, we will achieve a 20-30% adjusted EBITDA margin, which reflects our committed operational excellence and cost discipline. Third, with free cash flow consistently exceeding net income, our greater than 100% conversion rate underscores strong cash generation and working capital efficiency. To provide deeper insight on our financial trajectory later today, Dennis will walk through this financial model in more detail, including the key assumptions that underpin each of these levers.
I want to thank you all for your time this morning and throughout the day. I will be available for Q&A along with Rob and Dennis in the concluding session, and I look forward to answering your questions at that time. It's now my pleasure to welcome to the stage Reto Wyss to talk about our industry-leading AI. Take it away, Reto.
Thank you, Matt. Hello. My name is Reto Wyss. I lead the AI, R&D, and vision tool development at Cognex. I joined Cognex in 2017 through the acquisition of ViDi Systems, an AI startup I founded in 2012 in Switzerland. Overall, I've been active in the domain of machine learning, deep learning, AI for over 25 years. Ever since I joined Cognex, I've been working on getting exciting new AI technology into all of our products.
Today, it's my pleasure to talk about our industry-leading AI technology and how it permits us to compete with open-source ready available models. Because that is a question we often ask: how does Cognex compete with open source? Specifically, 20 years ago, that was with respect to OpenCV. OpenCV is a computer vision library that was originally developed by Intel and was then released in the public domain in the early 2000s. It is a collection of basic low-level vision tools that are used for—do you want to quickly check, maybe? Sorry, your mic is dropping out. Could you use this, please? Okay. Yeah, thank you. Does that work? All right. OpenCV is a collection of basic low-level vision algorithms, but to this day, it does not provide higher-level optimized tools like Cognex that provide capability, speed, and robustness beyond those basic vision algorithms.
Over the last couple of decades, Cognex has not only released advanced vision tools that dramatically improved the capability of rule-based machine vision, we have created tools whose name is synonymous for the best-in-class vision technology. In 1997, we released PatMax, a tool that still today is state-of-the-art in fast and accurate part localization. IDMax has set new standards in code reading. Hotbars is a technology that dramatically improves our capability to read 1D codes at very low resolutions. HDR + improves image quality such that we can also deal better with images with scenes that have a large range of contrast variations. Now, today, we face that same question again, but with a twist. How does Cognex compete against open source in the age of AI? It's a fair question.
The number of publications and models that have been released into the public domain ever since the early days of deep learning in the 2010s has increased dramatically. The last two to three years, this number has exploded since we are in our latest AI hype that got started when the world was introduced to ChatGPT and generative AI. In order to answer the question of how does Cognex compete with open source in the age of AI, we have to better understand what is the difference between generic computer vision and industrial computer vision. ImageNet is a large collection of images of natural objects and scenes, and it has been instrumental in advancing AI in vision research. Still today, it is basically the foundation of practically every AI open-source model out there. On the right, we see a typical example of an industrial computer vision task.
As you can see, these images look nothing like what you would see in ImageNet. It's not a natural image. It's highly task-specific, and oftentimes, for myself, I suppose for you just as well, we don't even really understand what these images show. They manifest complex textures, complex backgrounds, and oftentimes, the visual features and properties that are required to say whether that specific image is a pass or a fail are actually very small and very hard to discern. Not surprising that open-source off-the-shelf models don't do very well on these kinds of tasks. In the following, I would like to lay out the four main reasons or the four main ingredients that we actually need in order to make AI fit for the shop floor. It's accuracy, it's ease of use, it's efficiency, speed, and it's scalability. Let's start with accuracy.
In a typical industrial inspection application, we need accuracies of 99% or more. Let's imagine we have a production line that runs at about 600 parts per minute. If we would have only 1% of false accepts, that would mean that within one shift or eight hours, we would basically miss more than 2,800 faulty parts. That is clearly not where we can be. Let us now compare a Cognex model against a state-of-the-art vision language model on two different computer vision tasks. The first is a generic computer vision task. It's like ImageNet classification. As you can see, both models are somewhere in the mid-80s. That is probably good enough if a human interacts with a chatbot because the human, hopefully, is doing some sort of a plausibility check, or at least he should be.
For industry, for factory automation, we strive for 100% full automation. In that context, as I just laid out before, this is not viable. On the other hand, now let's look at the industrial computer vision task. Here are those same images again. I will now explain to you what we are seeing here. This is a frontal view of a plug. Basically, what we need to do is we need to inspect these two metallic pins in the center, make sure that they're not defective, make sure that they're properly aligned, and that they are in their respective slots. Naively, we basically can take that open-source vision language model and give it the following prompt: Are the two metallic pins in the center properly aligned within their respective slots? As it turns out, that first prompt did not work out very well.
The performance was very low, just barely above chance. After quite a bit of trial and error and prompt engineering, we found out that the model is actually not understanding what we mean with a metallic pin. It is only once we started to call these metallic pins metallic spheres that the model started to understand what we mean. We did not really get it much above the low 90%, as you can see here. In contrast, for the Cognex model, we used five samples of good parts, five samples of bad parts, and that was enough for the model to reach 100% accuracy. Next up is ease of use. Our systems are deployed into challenging environments, and therefore, it is very important that they are easy to train and easy to maintain. In order to train an AI model, you need to collect samples.
You have to take images. You have to label those images. You can imagine that the effort to do that is directly correlated with the number of samples you need. At Cognex, we've developed a proprietary technology we call edge learning that allows our models to train from much fewer samples, whereas reaching a target accuracy. We can learn from tens rather than thousands. As you can see on this illustration, that allows us to reach a certain target accuracy much faster. These models are much easier to train, much easier to deploy, but also much easier to maintain. Another constraint we face when deploying our models into factories is the often harsh and very demanding environmental conditions. You have to deal with dirt, with heat, with vibration.
Moreover, our customers would like to deploy these models into small, compact, and cost-effective packages. The ability to run these models on an embedded, rugged hardware is not just a nice-to-have; it is essential. Coming back to our comparison, here we see that our model, since it is several hundred times smaller, can run at line speeds on factory floors. Doing that on embedded hardware uses around 8 W of power. If we want to do the same thing with that open-source model, which is built from 7 billion parameters, we would need several large discrete GPUs running in a workstation or in a cluster thereof. Thus, we have optimized our model specifically to be able to keep up with modern production line speeds while keeping the accuracy and all of that in an efficient, easy-to-deploy industrial form factor. Moving on to scalability.
Even though edge learning allows our customers to train and deploy their models much faster, it's still a fair amount of work. Once I've trained a model on a specific production line, wouldn't it be great if I could take that same model and use it for the remaining 19 identical production lines that are in my factory? The issue is that if you take images from these different production lines, and you see here an example of EV battery cells, even though those lines are supposed to be identical, the images still look quite a bit different. You may say, "They don't look all that different, do they?" Given that we're looking for small defects and we want to be as close as possible to 100% accuracy, these differences actually do matter very much.
As you can see on the right, that normal model drops quite a bit in accuracy once you move from line A to line B. Last year, we released a new series of models, so-called robust models, that are much better in keeping up that performance. They can deal with these types of environmental changes, like changes in illumination, slight changes in camera position or camera angle, different focal planes, and they keep up that accuracy while moving from line A to line B. Similarly, in modern production environments, we move more and more towards high-mix, low-volume manufacturing. Similarly to the ability to move from one line to the next, we would like to be able to move from one SKU to another. Here you see an example of a cosmetic product.
It's a different product, but they all share the same issues or the same potential defects. Typically, we're looking at air bubbles under the label or at wrinkled labels. Now, instead of training for each of those products separately, what we would like to do is to be able to train on one and then transfer that from one product to the next. We have invented a new technology that allows our customers now to basically learn those defects, air bubbles, wrinkles, while not being specifically reacting to the change in the actual appearance of the product itself. Again, as you can see on the right, that allows us to keep the accuracy high when we move from one product A to product B, which is not true for the normal model.
Similar to our journey in rules-based machine vision, Cognex has been leading the way in bringing the benefits of AI technology to industrial manufacturing and inspection. We've created value for our customers by addressing the most pressing and most blocking issues. In 2022, we have released our first edge learning product, the technology, as I mentioned before, that allows us to train from as few as five to ten images. We've then brought that same technology to the 3D world by bringing AI to make easy-to-use, robust, but yet very precise measurement applications. Earlier this year, we've released a transformer-based version of our few-shot technology that allows us now to not only train from very few samples, but to actually keep up with the accuracy of models that have been trained on hundreds or thousands of images.
Also, this year, we have released our first AI-based DataMan barcode reader, the DM 290, where AI helps us to simplify and accelerate the setup process dramatically, but also at the same time, AI allows us to read codes that were previously not readable, something that I myself would not have expected to be possible just a couple of years ago. Let us now dive into this last example a little deeper. For an AI reader, ID reader, excuse me, the setup process is key to get highly accurate and fast and reliable reading results. In a rules-based traditional reader, I would say, as you can see here on the right, the screenshot, this setup process can be quite cumbersome.
You have to, specifically, if you're looking at codes like here to the right, to the left, the direct part mark or so-called DPM codes. For an experienced user, it has to explore a high-dimensional configuration space in order to find out what is the ideal camera position, distance, angle, what choice of illumination gives the best contrast, but also how to set up all the different parameters that control the actual reading strategy. Let me now show you how this is done with the new DM 290. In the first set, we set the part in front of the reader. We make sure that the code, in principle, is visible, albeit here it's not readable because we have these specular reflections and pretty poor contrast.
In a first step, we instruct the AI to do a rough analysis of the camera position itself in order to provide feedback to the user of how to position the camera, whether to move closer, whether to move further away. As you can see here, we're in about the right spot. We instruct the AI to start the tuning process. In order to do that, we cannot explore the complete configuration space, the high-dimensional configuration space that would take much too long. We have trained an AI model on many, many similar situations such that it can learn how to find that optimal operating point more efficiently. It is this process that you see here in action.
Also, we've introduced a new AI enhancer, which is a tool that allows us to clean the codes, but also to correct certain defects such that we can read codes that previously were unreadable. At the end of the process, the tool offers the choice of three different optimal operating points from which an experienced user can choose the one that they like most. You will get the chance to see this setup process in more detail in our upcoming auto product demo session later. To conclude, I would like to reiterate the four key ingredients that our customers ask to get AI successfully deployed into factories. We need 99% or more accuracy. The customer also wants to have something that is easy to train, easy to deploy, ideally from as few samples as possible.
Customers do not want their production line to be slowed down by an inspection process. We absolutely have to be able to keep up with line speeds. Also, customers would like to be able to take the investment they put into one training, one model, and be able to scale that to different sites, different lines, but also different product SKUs. We at Cognex have invented and released unique technology, tools, and products to address each and one of these requirements, nothing that we see in readily available open-source software. Now, with this, I would like to hand it over to Shirin Salim. She will be speaking, among other things, about how these different tools and technologies I have talked about here tie into the overall Cognex ecosystem. Please, Shirin.
Thank you. All right. Thank you, Reto. Good morning to all of you. I am Shirin Saleem.
I lead the software and applications engineering organization at Cognex. My teams leverage the cutting-edge AI technology you heard Reto talk about, weave it into our products, and make it available to our customers. A little about my journey. I joined Cognex in May of 2023, two years ago. After a long career in Amazon in the Alexa AI group, I worked on far-field speech recognition, natural language understanding, and building translation-powered experiences for millions of customers prior to coming here. Picking up from Reto, I'm going to talk about the foundational role AI plays in our comprehensive ecosystem, raising the bar in terms of how customers interact with our products, how they solve machine vision applications, and experience our products. What is this ecosystem? What makes it comprehensive? Why does it matter?
Our machine vision ecosystem is comprised of a complete lineup of products serving different personas of customers. They range from the novice line operator to the sophisticated machine builder. Our products support different types of applications ranging from 2D to 3D, simple to complex. They are built on a common software and hardware platform, which plays an integral role in standardizing the customer experience and incentivizing our customers to adopt more of our products and not migrate to vendors for new applications. AI technology is an integral part of our portfolio. We offer both deep learning technology that delivers the highest accuracy for complex and data-rich tasks, as well as edge learning technology for simpler tasks that can easily be trained or adapted to a customer's application with just a few images directly on the device.
Today, I'm very excited to announce that we have a new pillar in our ecosystem, a cloud offering that makes it even easier to incorporate advanced AI models into our product lineup. We issued a press release on that offering after markets closed yesterday, in case you haven't already seen it. I'll talk more about that shortly. Our competitors, both established players as well as the newer startups, may offer one or more of these pillars, but none offers an ecosystem that is as comprehensive and rich as ours in advanced machine vision. Let's talk a little bit about what makes our ecosystem attractive for our customers. Let's start by imagining a day in the life of a sales engineer visiting a customer. He or she will typically introduce the In-Sight 2800 series of embedded vision sensors. It's designed to solve common 2D applications.
The sales engineer, after talking to the customer, may realize that the application requires more speed and more capabilities than the 2800 series can offer. So what does the engineer do? He or she can then offer our advanced In-Sight 3800 series of vision systems that has the same easy-to-use tools in the same software environment, but with better performance as well as higher resolution and lighting options. This provides a more effective solution for customers that have applications that require a higher level of performance. Now, in other cases, a customer may have a new application, one that requires precise 3D measurements, such as inspecting the height of small pins on a PCB. The sales engineer can then recommend the AI-powered In-Sight L38 on the very same software platform, but offering advanced imaging to capture detailed 3D images and analyze depth information.
The customer, already being familiar with the vision tools and software, can transition more quickly to the new product, transferring all of their prior work. They do not need to invest in extensive retraining or reconfiguration or reintegration with their PLCs. The standardization of cabling and mounting allows them to follow a familiar setup process. Cognex application engineers, whom customers have developed strong relationships with and who have an in-depth understanding of their use cases, are still highly effective in providing solutions tailored to their needs, even with a new product. Our journey towards a common platform actually started several years ago with the goal of improving our own internal efficiency. Over time, we found that this approach has become an important selling point for us, saving time and cost for our customers.
You will hear more about how our sales force is organized to capitalize on the benefits of both the flexibility as well as the commonality of our ecosystem from Carl in the next section. Let's talk now a little bit about the role of AI in our ecosystem. We offer both deep learning vision tools and edge learning tools across our products. Let me reiterate again, deep learning tools are designed for complex applications that require high accuracy, but also thousands of images to train. Edge learning tools, they're designed to be quickly trained on an embedded platform with just a few images, but they work best for applications that are not too complex. We offer these options across our embedded and vision software products. We allow our customers to make the right choice depending on their application.
As we talk to our customers, we found that we have room to make the experience even better for them. Customers do not want to be making that trade-off between accuracy and ease of use. They want both. If they do need to make that transition from edge learning to deep learning, they want to do it seamlessly without having to redo the work they previously did. I am excited to announce OneVision, the first fully released cloud-based no-code machine vision AI training platform. With OneVision, our customers can seamlessly integrate state-of-the-art AI models you heard Reto talk about into their vision jobs, and they can access advanced models that offer deep learning performance with the ease of use of edge learning. OneVision is available on the In-Sight 3800 and 8900 series now and will launch on more products in early 2026.
With OneVision, you can think of our vision systems getting smarter. Customers can train AI models to solve an application. They can then combine it with highly precise rules-based tools of their choice and deploy it across our full family of products. Customers could get the benefit of a cloud-based product when they're designing their machine vision system. They can combine that with all the benefits of edge computing, including low-latency inference when they deploy these models on the In-Sight products. Some of the benefits of the cloud, what are they, right, for when you're designing a job? It offers faster training of AI models on multi-GPU clusters. On the cloud, you have access to advanced AI models and architectures, and we provide built-in tools to distill them to run on Cognex's embedded products.
You can do cross-site or cross-production line management of projects, which allow for easy collaboration and coordination. OneVision simplifies the entire lifecycle of data curation, AI model training, validation, optimization, and deployment on our vision systems. Think of the journey starting and ending at the edge, and it integrates with our In-Sight product portfolio. Let's revisit the pin example that Reto introduced previously. To recap, the vision task at hand here is to inspect the pins in the center and classify them as good if they align perfectly in the circular hole and bad if they do not. Now, the complexity of this task increases when you have lighting and part variations across lines, and it could take hours to configure multiple AI models for each line, each of them trained on the edge with a few images capturing some diversity. Let's see how a customer would solve it with OneVision.
Step one, images of the pins collected from one or more lines are uploaded to a common OneVision project with a push of a button. Step two, using a few labeled images, the customer trains an AI model to locate the pins. Now, labeling of the data is fast and efficient thanks to different modes of AI assistance offered in OneVision, including system-in-the-loop labeling as well as point-and-click labeling. Now, once the model to locate the pins is trained, the customer then trains another model to classify the pins as good or bad. Different types of deep learning models are offered, one that works with just a few samples or more advanced models that use more images for training if they are available. At all stages, the customer has access to built-in assistance to highlight any corrections or opportunities for further optimization.
Finally, once the models have been trained and validated to be robust across part variations and lighting conditions, they can be deployed to all cameras across multiple lines or sites. It is easy, it is fast, and it is scalable. With OneVision, the cycle of data collection, training, validation, deployment, and monitoring of AI models spins much faster across the entire Cognex ecosystem. OneVision offers benefits for customers in all of our verticals. Our customers in consumer electronics typically employ vision engineers with years of experience. That is because they need to design complex vision jobs. With the advanced AI models we have available in OneVision, they can configure the same job with operators who have just days of experience with vision systems. In logistics, there is this challenge due to the diversity of different items that are encountered on the conveyor belts.
You have to train the models on different types of items for it to be robust. With the powerful compute of the cloud, it takes just hours to train new models on unseen and diverse items, significantly reducing the lead time to adapt to new products. Our automotive customers, they typically need to deploy deep learning models across multiple lines and sites, each with varying production conditions. What previously required hours of repeated effort for each site is now seamless with OneVision's ability to train and deploy across multiple sites on the fly. To reiterate, we are pushing the envelope of AI technology for machine vision. We have a strong foothold in the embedded machine vision market with our comprehensive ecosystem. With OneVision, now we can go even faster in embedding these powerful AI tools in the products in our ecosystem.
Overall, OneVision strengthens our ecosystem and positions us very well against our strategic objective of being the number one in AI technology for industrial machine vision. Do not just take my word for it. Why do you not see it in action and hear about it directly from our customers?
Yeah, I mean, we did not really have to do anything, honestly. We just put it in there and click next a few times, and here we are. Cognex products are designed to be easy, but that does not mean they are simple. Building on decades of leadership in advanced technology, Cognex AI helps our customers solve their complex machine vision applications fast. The DataMan 290 barcode reader uses advanced AI technologies that guide users to make setup easy and push read rates up by rapidly decoding challenging or damaged codes.
Cognex AI also makes it possible for In-Sight vision systems to solve challenging visual inspections. With just a few example images, any level of user can set up and deploy in minutes. If customers have questions, we can help in multiple ways. With curated self-service resources, over 500 support engineers around the globe, and digital solutions for seamless interactions over multiple channels, Cognex ensures that customers succeed. The constant pace of innovation that we are seeing from Cognex, we can depend on something that is better, faster, cheaper. I would say the main advantage of using Cognex is the ease of use of the products. The software, the setting up of the tools is one of the easiest I have seen by far in the industry.
That's one of the things that's really excited me about the way Cognex has formulated their product is it really brings to a customer solutions that haven't been there before.
All right. With that, I'm going to pass it on to Carl Gerst to talk about sales as a competitive edge.
Great. Thank you, Shirin. It's wonderful to see so many familiar faces here today. For those of you I haven't met, my name is Carl Gerst, and I've been with Cognex for 25 years. I began my journey at Cognex as the first product manager for our In-Sight vision systems. In 2004, I helped launch our ID products business, which I led for 15 years. What started out as a small initiative grew to represent nearly 40% of our overall business. A major driver of that growth was our strategic move into logistics. That.
More recently, I've worked closely with Matt and several others on evolving our product development teams, which has helped build the ecosystem that you just heard about from Shirin. In the middle of last year, I transitioned to lead our sales teams. It's been an exciting change for me. It's bringing me back to what I enjoy most, which is working directly with our customers to understand their unmet needs and finding ways to better serve them. That's exactly what I want to talk about with you next. Before we dive into the next section, I'd like to briefly reflect on what we've covered so far with Reto and Shirin. We explored how our industry-leading AI capabilities and our comprehensive product ecosystem are not just nice to have, but how they're strategic differentiators for us.
These two elements are critical components of our go-to-market strategy and delivering a best-in-class customer experience. As Matt mentioned earlier, we have evolved our go-to-market strategy by making targeted investments in our sales force. This is enabling us to engage more deeply with our largest tech-savvy customers, and it is also enabling us to unlock new customers with easy-to-use and easy-to-deploy products. We believe that customer acquisition and customer experience are not separate efforts. We see them as part of a powerful self-reinforcing flywheel. If we take a closer look at that flywheel, you will see at the center of it is our goal to double the number of customers over the next five years. This flywheel is powered by our sales force transformation and our goal of providing the best customer experience in the industry. This helps us build sticky, long-term relationships. Surrounding this are three key actions that drive momentum.
First is specify your vision. This is where we work with our customers to understand their goals and define the outcomes that they want to achieve. Second is deploy your vision. This is where we bring those goals to life, and it is tailored through solutions and seamless implementation. Third is support your vision. This is where we ensure long-term success through ongoing support and partnership at every step of the process. A key enabler of this model is our direct sales channel. It allows us to engage with customers across the journey from the initial contact to ongoing support, ensuring consistent, high-value experience at every touchpoint. Let's take a closer look at our direct sales channel. We have two sales teams, one that is focused on logistics and one that is focused on factory automation. Both of these sales teams engage directly with our customers to understand what their needs are.
The way that our products and solutions are delivered may vary based on the project. For example, if we're working with a customer and they have an existing line, they may be working directly with us to be able to buy an upgrade to that line with our ID and vision systems. That same customer may be looking to expand capacity and add a new line. In that case, it would not be uncommon for that customer to buy those products through one of our partners, such as a machine builder or a systems integrator. The Cognex ecosystem that you just heard about from Shirin plays a critical role in accelerating the success for both our sales teams and our partners. I'm excited about the launch of OneVision, most notably the ability to bridge between edge learning and deep learning. Why?
Because it enables faster development, it enables faster execution, and it also provides a significantly better customer experience. Additionally, our AI-powered products that you heard earlier about from both Reto and Shirin provide user-friendly, guided workflows. That is empowering both our partners and our customers to move faster. We are focusing our sales team on three themes, each with a dedicated mission. We have a dedicated team that is focused on market creation and expansion. These teams are focusing on acquiring new logos and deepening our reach at existing customers by showcasing products like the DataMan 290. These teams collaborate closely with a more senior sales team who concentrate on market penetration. This team is focused on increasing share of wallet within existing accounts by leveraging the full Cognex ecosystem. They also play an important role in supporting new logo accounts, particularly when more sophisticated vision is needed.
We also have a team of sales engineers that's focused on supporting value-added partners and helping them move faster. This would include machine builders, systems integrators, as well as our automation solution providers. They coordinate efforts between the partner and the end user, and they capitalize on our unified ecosystem and seamless integration capabilities. Our investment in our sales force has expanded our sales coverage, and it's allowing us to better reach previously underserved customer segments. At the same time, our commitment to AI-enabled innovation has led to the developments of market-leading products with feature capabilities such as OneTouch AI-powered image formation, AI-guided positioning, tuning, and decoding. Why are these features so important? Because it simplifies the qualification and deployment process with our products. You're going to hear more about how we've enhanced our sales reach combined with these cutting-edge AI capabilities in the upcoming packaging product demonstrations.
We see a strong opportunity to grow market share at existing accounts and drive sales by leveraging the power and flexibility of our expanding ecosystem. Tech-savvy enterprise-level customers have long appreciated our ability to solve their most complex applications. What that has often required is them to use a mix of products that often felt disconnected. Our unified ecosystem changes that, streamlines everything from development to deployment, resulting in better results and a better overall return on investments. This means that customers can start with our 2D systems and seamlessly scale to an AI-powered system when 3D measurement data is needed. As you heard from Shirin, our OneVision platform now enables customers to move from using edge learning to deep learning for inspection tasks for more demanding applications, and all while staying within the same ecosystem.
This integrated approach, combined with a full Cognex ecosystem, is resonating strongly in key markets such as automotive and logistics. You are going to get a chance to see that after these sessions in our automotive product demonstrations. One of the key advantages of the ecosystem that we are building is how it accelerates our sales team and our partners' efforts by offering solutions that are easier to quote, easier to order, easier to install, easier to commission, and easier to support. We have long been recognized for our leading-edge technology, but it has not always been the most straightforward to implement or to maintain. By transitioning to standardized solutions with guided workflows, it has simplified the process for our partners, it is significantly enhancing the customer experience, and it has improved operational efficiencies. Our focus here has been on logistics and consumer electronics.
I can tell you firsthand from spending a lot of time in the field this year, the positive feedback that we're getting from both partners and our customers is incredibly encouraging. You're going to see this reflected in the upcoming logistics product demonstrations. I now want to turn to customer experience and take a closer look at what we mean by specify your vision, deploy your vision, and support your vision. First, specify your vision. We're empowering our customers, providing the tools, guidance, and flexibility they need to define their own path forward. We've transitioned from expert-driven setups to self-guided automated workflows. This is enabling our customers to reduce specialized resources, and it's improved overall scalability and is providing a stronger return on investment.
Through a connected product ecosystem, we're beginning to deliver a more personalized experience with features like guided learning and AI-powered agents that can answer application-specific questions. What's this doing for us? It's enabling our customers to work more independently and autonomously. What that, in turn, is doing, it's enabling our teams to engage in more meaningful and consultative interactions with our customers. What excites me most about the evolution of our product ecosystem that was highlighted by Shirin is, rather than presenting a disconnected mix of products to our customers, we now can offer a comprehensive integrated ecosystem. This is allowing our customers to more quickly specify our solutions for new applications. It's accelerating market penetration within existing accounts, and it's also helping us attract new customers as well. Second, deploy your vision.
We've streamlined the deployment of our vision systems, and this is delivering confidence with our sales teams, our partners, and our customers. On the software side, we've evolved from delivering custom solutions that were built on legacy architectures to app-based models that are based on scalable software architectures. This shift has enabled the creation of guided workflows that you're going to see in the demonstrations. It's not only efficient, but it's also designed to scale seamlessly. On the hardware side, we've moved from bespoke, resource-intensive solutions that were often tailored to individual customers to standardized hardware offerings that address a wide range of market needs. This has led to faster deployments and significantly improved operational efficiencies. Together, these advancements are driving the automation of workflows, which is resulting in streamlining installations, reducing the need for skilled labor, ensuring seamless integration, and optimal performance across the board. Third, support your vision.
Our investments in a digital platform are transforming the way that we're connecting with our customers. It enables us to deliver seamless integration and drive success through every stage of the customer journey. By being digitally connected, we can now offer proactive, real-time support through our MyCognex customer portal. This allows us to accelerate issue resolution, and it's helping our customers maximize uptime. These digital capabilities allow us to not only connect with our customers, but it's beginning to enable us to connect directly with our products. That is enabling us to do things like remote diagnostics and automated triage when issues arise. All of this is helping us improve our responsiveness with our customers. These digital services are enabling us to deliver tailored support solutions that address a wide range of customer needs.
We're still early in our digital journey, but the impact of enabling connected support at every step of the way has already surpassed my expectations. I'm going to end where I started. We believe that customer acquisition and customer experience are not separate efforts. We believe that our direct sales channel, along with our comprehensive product ecosystem, are strategic differentiators that enable our goal of providing best-in-class customer performance and ultimately doubling the number of customers that we serve. I'm going to ask Reto and Shirin to join me on stage, where we'll be happy to answer your questions. We're just going to set up some chairs here and we'll get started.
All right. Before we jump into questions with Carl, Reto, and Shirin, please note that we have 25 minutes for this Q&A session. We ask that you kindly limit yourself to one question.
We have more time this afternoon at the end of the day for additional Q&A. Please wait for the mic runner to hand you the microphone before you begin your question.
Hi. Good morning, everyone. Thanks for the great detail today. When we think historically about machine vision and sort of the hurdles to adoption, I guess we think about a big, expensive piece of PC hardware that had to be off to the side of a production line. Is cost no longer the main issue, given all the advancements that you've had in the technology and being able to put it into a smaller form factor? Is it really just about education and sales coverage, or how would you think about what the biggest issues are today for customers in terms of adoption of machine vision more broadly?
The question is whether cost is the driver or whether more broadly there's other elements that are important to it? Right. Exactly. Yeah. For those of you that were at the customer session last night, I think you heard quite a bit about this. I mean, I think cost is always a driver, right? What else are they thinking about when they're thinking about implementing a vision system? They're thinking about the cost of the product. They're thinking about the cost of integration. They're thinking about the cost of long-term support. I think more and more for us, what we're looking at, I think we've often in the past focused on the technology and providing the best technology. What we see is that we can help our customers lower their overall total cost by providing a significantly better customer experience.
I think the ecosystem that you heard about from Shirin, as well as the tools that you heard about from Reto, is helping us really lower the customer's overall total cost of ownership, which often we think is being beneficial by enabling us to sell products at a higher price.
All right. That's helpful. Just to first, Rory, and just a quick one on the OneVision cloud offering. I mean, is the pricing model different versus selling just a piece of hardware today? Now, is there a subscription component when you've got a customer that's being plugged into the cloud offering?
Yeah. The question is whether there's going to be a different pricing model with OneVision. What I would say is that we're not going to talk specifically about that today.
Our focus on the second half of this year, if you saw the press release, is we're really going to be focusing on select customers. A key thing that we're going to be focusing on is really the customer experience. I've seen this play out in real time, right? Where customers, some of the—I was out with customers earlier this year, and the feedback was, "Your edge learning tools that you've developed and introduced over the last couple of years is the best thing that they've seen from Cognex in a long, long time." Right? What you see is those customers will hit at some point in some applications, will hit a point where they need more performance. Today, they're having to jump off into a different platform. With OneVision, we're able to keep that in, meaning they can stay within In-Sight, right?
We're introducing this with the In-Sight 3800 and the In-Sight 8900. That customer can stay within In-Sight and update the model on that system. I would say we're looking at the pricing, right? We're looking at how that's going to be priced. We're going to use what we learned from the second half of this year to influence that as we go into next year.
Hey. Good morning. Joe Ritchie from Goldman Sachs. Thanks for all the details today. Carl, my question is for you. It's a multi-part question. As you think about the 3,000 customers that you added last year, I'm curious what portion of the customers came from new logos versus share of wallet with existing customers. Part two, as you look to expand and double your customer base over the next five years, how do you envision that shifting?
I'm curious, what does the cost element of this equation look like in terms of trying to penetrate and double your customer base?
All right. Let me try to break that part a little bit. Three parts. Maybe the first part, and you'll have to maybe help me go through the second and third part. I think your first question is on the 3,000 customers that we added, what percentage of them are new logos versus what percentage of them maybe would come from existing? I'd say those are new logos customers, right? What we're trying to do is, with expanding our sales force, the team that I talked about that's really focused on market creation and expansion is really focused on calling on new customers and attracting new customers.
They work very closely with a second team that's really focused on market penetration. What we would see is the ecosystem that we're developing helps us in both cases. It helps us go deeper with existing accounts. I think Shirin kind of described this really well in her presentation. Oftentimes, we're in there with an In-Sight 2D system. For 3D, we're all of a sudden competing all from scratch, right? Because it often feels, even with our competitors, it feels like their products are often coming from different companies. If you went back a couple of years at Cognex, if I went into a customer, it would feel like our 2D systems, our 3D systems, were all coming from different customers. A common ecosystem, what does that mean, right? They know the software. They know the integration. They built their HMIs, right?
That gives us a big leg up when we're looking to penetrate that account with new applications. I would say to your initial question, those 3,000, those are really new logo accounts that we added. Maybe.