Presentation. My name is Max Mock, and I'm the research analyst here at William Blair who covers Certara. We're pleased to be joined this afternoon by CEO William Feehery. Before we get into the presentation, I want to mention two things, or I need to mention two things. First, the breakout session will be held in Genie B on the second floor immediately following this presentation. Second, I am required to inform you that for a complete list of research disclosures or potential conflicts of interest, please visit our website at williamblair.com. Again, very pleased to have Certara with us here today. With that, I'll turn it over to Bill.
Thanks, Max. Appreciate it. Now you can see and hear me. So that's good. All right. So Certara, we are focused on a problem, which I think is a big problem that we all care about. And that's the overall issue that—can you guys hear me?—that drug development needs a new model. Despite the fact that billions of dollars have been poured into pharmaceutical research and development, it is still the case that almost 90% of new medicines that enter clinical trials fail. That statistic, if anything, has gotten worse over the last generation. When you add up what that really means is that the cost to develop a new drug, most sources put it at $2 billion-$3 billion. Most of that $2 billion-$3 billion is spent not on developing the drug that gets approved, but it's on all of the tries of things that don't get approved.
When you kind of look at this, you say, "Hey, there ought to be some changes that can be made to the system." If they are, this is a global system that affects everybody. It could really change the face of pharmaceuticals and the face of human health. We have done a little bit of modeling on this, and we said, "What happens if we could just make small changes that would have a big effect?" What we did was we said, "We are not saying we can only make a 3% improvement.
If you just looked at every phase of drug development and you made a 3% improvement in the probability a drug would get through that phase, you'd rapidly come up with a pretty big number in how much you can save per approved drug: $300 million. The way that Certara [audio distortion] puts itself on doing this is we work on what's called model-informed drug development, or what we call biosimulation. This is basically what we're doing: we're building computer models of how diseases [audio distortion] work in the human body and how drugs work across populations of humans that are, in fact, quite different in many ways.
We have built a biosimulation platform which enables our drug development customers to save years of time, reduce costs, and generally also increase scientific understanding, which is maybe an equally important part because we would like to know why the drug works, and we would like to be able to explain it to the FDA, and we would like to be able to improve as we go forward. Certara today, we are about 1,600 employees worldwide. We cover pretty much everywhere in the world where we have significant active drug development activity going on. We have a pretty extensive customer base. Given the breadth of the software and the services we have, we have pretty much all of the active entities that are developing pharmaceutics in the world: about 2,400 customers. We have all of the big pharma customers as long-term clients.
If you look at the success stories we've had, if you look over the last 10 years, we've traced over 90% of the drugs that have been approved by the FDA have used our technology in some way. If you want to look more specifically, we've looked at the filings of over 120 novel drugs where we can see specifically label claims which have enabled those customers to have avoided expensive human clinical trials in getting those drugs developed. We have a lot of success. Our company is profitable. We operate in the low 30% EBITDA margin. We're investing heavily in software growth. We have a healthy balance sheet. We occasionally acquire companies to continue to grow.
Although we've been around a while in one form or another, our company was formed over a number of years of acquisitions. We believe that there's still a very healthy opportunity to continue to grow this as we go forward. Let's talk a little bit about what we do. We have a biosimulation platform. If it is used in the recommended way in a drug program, what we do is we start as early as possible. The process is one in which what we do is we build a model of that drug given all of the available information at that time. In the beginning, it might be very little information. It might be just some information about the chemical structure or information about a drug that has been approved in the past that might be sort of similar.
As we go forward, we're accumulating information from lab and clinical trial data, and our models get better. At each stage, we're able to basically predict what we're doing is we're using these models to predict what's going to happen to the next stage. We refine the model, and we move on. People often ask me, "Well, what's your exposure to different stages of drug development?" The answer is, if this is used right, you start in the beginning and you go all the way to the end, and you're exposed to all of them. Now, you're answering different questions as you go along. In discovery, we're answering basic questions: what's the target we're going after? And among all these candidates, which one should I select would be kind of a typical one.
As we go into preclinical, the questions that we're often answering are first in human dosing. We have animal trial data. We're going to go into a phase I clinical trial. We want to basically have a dose range where we're going to basically hit that active area. We're not going to miss it or something like that and waste our very valuable data points testing in a regime that would either be toxic or not work at all. As we go into early clinical, we're answering questions like inclusion and exclusion criteria. Are there going to be populations of people that respond differently to this? Maybe we want to exclude them. Maybe we want a different dosing, different treatment regime.
In late clinical, we make very common use for what we're doing is things like drug-drug interactions where, for a lot of drugs, you can use simulations to avoid a very expensive late-stage clinical trial. That's pretty commonly accepted by the FDA. You're answering different questions, but basically, you're still starting the beginning with a model. It just gets more sophisticated, and you can answer different questions as you move along. Is this a study that was recently published by Pfizer? It has nothing to do with us, but we thought it was very interesting where they talked about model-informed drug development.
They looked over a number of years at all of the drugs that Pfizer had worked on, not just approved, but the ones that were worked on by Pfizer and came to the conclusion that this saves 10 months and $5 million per program in their hands. Now, this is a tough study to do because some of the biggest impact we have is actually on telling people to stop developing a drug that should not otherwise be stopped. You can understand the difficulty in sort of measuring your impact if you're trying to compare something that the person didn't do as they went forward. Still, I think what you're seeing here is recognition by the industry that this is accepted, has a big effect, and we can find other examples. This is just a recent one that came out over the last couple of months.
You can see other examples in the industry of how this is being adopted as well. The graph on the left basically shows the number of publications over the last 25 years in biosimulation. You can see an exponential curve in basically not only the interest, but the development of the area. The one on the right's kind of interesting as well. The FDA is a big piece of our story. Nobody does anything in drug development unless the FDA approves of it and ideally encourages it. This is a plot of the guidances that the FDA has put out, which specifically mention various uses of biosimulation that they would approve or expect in drug development.
You can see that's been rising over the years as the technology has gotten better and as the FDA has gotten more comfortable with the way it's used and what's in the models. We have tried to address this in thinking about the addressable market. It's obviously not an easy thing to do. We have examples of customers that do everything we recommend from start to finish. We have many, many more examples of people that come in for some part of that process that I talked about, but maybe not all of it. We made a few assumptions here to kind of give an idea of our idea of how big the TAM is here. We kind of did the same thing I talked about before, but assuming like, "What if we could just improve the success rate by a small amount?
We could capture 20% of the value we deliver. You kind of multiply that by how many drugs the FDA typically approves and how many NDAs there typically are in a year that, assuming that does not change, we come up with a market TAM of about $4 billion. Other companies externally have done similar studies from different approaches and sort of come up with similar numbers. This is a significant market where I think we are probably the most significant pure-play biosimulation company that is out there. We are obviously nowhere near $4 billion. I would say two things to that. One is there is a long way to go.
Second is there's a lot of activity here that's happening within our big pharmaceutical clients, all of which have very large groups that are doing this type of thing, but all of which are also dependent on the software and the technology that we're providing to the market. One of the other tailwinds that's happened to us recently was a recent announcement by the FDA that they plan to phase out animal testing. They specifically mentioned monoclonal antibodies. And we've been working on this for years with quite a number of our clients. It wasn't an accident, I think, the FDA picked monoclonal antibodies. It's a big category of drugs. It's probably 20% of what's been worked on out there. The models are widely used and accepted for dosing, so they're out there. It makes sense.
Frankly, there's been a lot of questions about the utility of data where we're giving a human antibody to a non-human model. In 2022, Congress came out with the FDA Modernization Act, which enabled the FDA to do this. They did basically nothing. The new administration has come in and made this a priority. We see this as a tailwind as we go forward, given that these models are right up our alley. Now, I'm going to talk a little bit about within our biosimulation platform, we have a number of products. They all have come forward to form a biosimulation platform. As I said, we span the drug development stages from discovery through clinical. In discovery, we have a ChemInformatics platform where we're using that to predict the chemical properties of a drug candidate. We acquired that recently through an acquisition called Chemaxon.
I'll talk about that product suite in a minute. Our biosimulation sort of flagship product is called Simcyp Simulator. Within biosimulation, I talked a little bit about how we can predict the next phase. We also have a suite of software where every time data comes in from clinical trials, we can fit that data to a model. This biosimulation and the analytics sort of work in tandem with each other. Our product is called Phoenix. Lastly, but certainly not least, we've invested a lot in data standardization and validation with a product called Pinnacle 21. The amount of data that is flowing through a big pharmaceutical client on a single drug, they're coming from, in some cases, hundreds of CROs. We need access to that data in a standard form to apply biosimulation.
Our clients are very interested in standardizing it so they can report it. They can make sense of it and report it to the FDA. That's another big piece of what we provide. I'll talk a little bit about these four main products briefly here. Let me start with our flagship product, Simcyp Simulator. This product is used, as I said, to model the kinetics of how drugs work in a human body. We are modeling all of the organs of a human body, but it's more than just one human body. The way this thing works is we're modeling what happens across populations of humans that vary. Populations of humans vary, obviously, by age, sex, weight. They vary by genetic differences, for example, some sort of genetic change or genetic mutation. They vary by racial characteristics, things like that.
They vary by comorbidities, people with kidney disease or liver problems and things like that. Also, for example, children and babies and neonatals are also quite different. We are interested in not just modeling the kinetics of a drug in one body, but when this gets out into a big clinical trial around a population, what is likely to happen and where the edge case is likely to be. This has been around. We started this in the early 2000s. It gets updated every year. It is quite a significant piece of software. It is quite unique in the world. We have found, just looking through drug filings, label approvals in over 120 novel drugs where companies have gotten an FDA approval specifically using the software to avoid clinical studies. Clinical studies is basically where all the money goes in pharmaceutical development. All right.
Just to give you an idea of one of our clients published case studies. I'm not going to go through all of this. This was Novartis for a drug called Scemblix. You can see the way this typically works. As I said, you start early on. You're doing some early modeling. This drug took roughly 10 years to be developed. It was a very interesting drug with a whole bunch of nonlinear components to it and also additional concerns around drug-drug interactions and different doses. They specifically in this paper talked about how the use of Simcyp enabled them to avoid multiple clinical trials for various questions that they wanted to either prove to the FDA or the FDA was going to ask them. They cited that they had avoided more than 15 human clinical studies.
Now, considering that this started 10 years ago, and that 10 years, if you went forward today, the technology is 10 years better, I think we could do even better if we were starting a drug like this today. Our drug for analysis is called Phoenix. This is used by the FDA. It's used by a number of other regulatory agencies. It has by far, I think, the largest market share of an analytics tool in pharmaceuticals. I won't go into this a whole lot, but the idea is generally you get clinical trial data. You want to fit your model to that clinical trial data so you can move forward and by a simulation. This is widely used. Although this product has been around for a number of years, we are making some very significant advances in AI to improve this product.
We're still adding a lot of features and sort of moving people up in terms of the number of seats and the value we're providing that we can capture in our pricing. Pinnacle 21 is our data standardization product. We bought this company in 2021. What I would say about Pinnacle 21 is this started out with the FDA. They wanted everybody to submit clinical trial data in a standard format. This format is called CDISC. The FDA uses Pinnacle 21 basically as sort of a gate. If you submit your data to the FDA, they use Pinnacle 21. Your data either passes a standard or it doesn't. Effectively, everybody who submits a drug to the FDA uses our software. When we bought this in 2021, this was maybe 75% penetrated for that particular use. It's probably nearly 100% right now.
If that was all the story was, it would still be a pretty good acquisition for us. However, we saw a bigger opportunity here. We say, "Why did the FDA want this data in a standard form?" They have the same problem that any large pharma company does, acquiring their data from their data suppliers, from labs, from CROs. We have added a lot of features on there where now this software is attacking a much bigger market in terms of looking at our pharma customers, how they're bringing all these data sources together. They're standardizing it and validating it and getting it to a form that basically can be analyzed and also used in biosimulation. Last but not least, I'll talk about our discovery product. This is our most recent acquisition.
We set out a couple of years ago saying, "Hey, one of the problems we have in Certara is that people are bringing us drugs after they've already made the molecule selection idea." We can only do so much. If you've already selected your molecule, we can tell you how to design a trial. We can tell you maybe you should stop. We'd like to get at this issue around, "Could you have made a better selection process?" We did a multi-year look at the drug discovery space, and we acquired this company called Chemaxon. What I'd say is this is a key piece of software in what's called the design, make, test, analyze cycle in discovery. The core software that we acquired is a chemical search engine.
This enables us to search databases of millions of chemical structures looking for the right one to bring forward. We are integrating this in our wider biosimulation platform this year. I would be remiss if I did not talk about AI, but we are, obviously, given what we do, a prime candidate to insert AI into biosimulation. We acquired a company called Vyasa a couple of years ago, which was very fortunately timed to give us a really good head start as LLMs became sort of on the map in terms of what they can do. What we have done is a couple of things. I am not going to go through this entire chart, but the way you can look at this is one is most of our biosimulation products have features which can be enhanced with AI, right?
We can do things like we can make them easier to use. We can make the kind of the very sophisticated experts that use this kind of software much more productive. We have launched specifically new products around, for example, reporting. One of our newest products, CoAuthor, will basically let you take all of the data and biosimulation, all of your clinical trial data, and write an entire draft of an NDA using that and saving lots of experts time. We've also got numerous other features around using predictive analytics to do things that we were unable to do with sort of the traditional mechanistic modeling that we do in biosimulation. If this was all we were doing, this would still be pretty good.
There's lots of we haven't disclosed how much we make in AI specifically, but let's just say it's already in the millions of dollars of revenue for us and growing nicely. Beyond that, we can do some even more interesting things. We have a product called Certara Layar. This is kind of an infrastructure product behind AI. What Certara Layar lets you do is you can basically plug in data sources that are proprietary or public. You can do this in a secure manner. You can do this without taking the data out of a database and giving it to somebody like Certara or an external party. You can create a custom AI for every one of our clients using exactly the data they want. We're using this in Certara to basically pull together our biosimulation platform.
Our customers are using this as basically an easy way to start to pull together these sort of custom and very sophisticated AIs that they can do, given in some cases the 50 or 100 years of data that they've got to work with that they don't want to get out of their company. M&A has been a piece of the story here in Certara for a long time. The company, it's hard to say exactly when we were founded because we did a lot of tuck-in acquisitions in the beginning. Since our IPO five years ago, we've done three significant acquisitions: Pinnacle 21, Applied BioMath, and Chemaxon. Pinnacle 21, when we acquired it, was a great company. Let's just say it's well more than double what it was when we acquired it. Applied BioMath [audio distortion] pharmacology player, we are now the largest players in that very growing area.
Chemaxon, as I mentioned earlier, is our discovery play. All of these companies, I think that we are quite disciplined in our acquisition criteria. We were very interested in the strategy. Obviously, the financial aspect of it is very important to our shareholders. We have been very successful with the integration and continue to grow these companies as we go forward. They are a big piece of our strategy as we move forward from here. I will just turn last, but not least, to our financials. Certara's growth profile over the last 10, excuse me, over the last three years, our software CAGR has grown. Our overall CAGR is 10%. Software is growing faster than services. Some of that is by design. We are specifically, I will talk about this in a minute, but part of our strategy is to increase our investment in our biosimulation software platform.
Certara is we report in two segments, software and services. The reason we do this is if you'd like to buy the software from us, many of our larger clients do. They use it internally. They build upon it. They're quite dependent on it. Our smaller clients, they often don't have internal biosimulation groups. So we need people with a lot of drug development expertise where if you don't want to buy the software and hire a group, you can simply hire us and we'll do the project for us. It's sort of a way of accessing the smaller size part of the market. You can see our bookings and our revenues have grown nicely throughout the year. As many of you know, there's a lot of things going on in the healthcare market which are a little bit difficult right now.
We are continuing to grow through it on both software and services. I'll show you a little bit about that here in a second. We are explicitly moving our revenue mix to software. Some of that is it strategically makes a lot of sense. Our software has a very high barrier to entry around it. The technology has gotten to the part it makes sense. Our inorganic strategy has given us a platform view of this whole thing. Part of our services group is under strategic review. We're considering whether we want to keep that. Right now, we're in the middle of a process around that, assuming that we don't. We'll be more like 50/50 software services as we go into the future. Lastly, we are explicitly investing in software through the current market situation.
You can see our software R&D spend increasing every year. It will increase again this year at a rate that's faster than sales. The reason for that is simply because the opportunity here is so significant to create an integrated software biosimulation platform. In summary, we're the leader in biosimulation. Biosimulation has got a lot of tailwinds. AI, it's not the whole story, but it certainly is helping us a whole lot, and we're investing in it. At the end of the day, we're a key piece of the story here about how you can improve pharmaceutical development worldwide and save years of time and risk, which benefits everybody. Thank you very much.
Thanks, everyone. We're going to go ahead and leave it there and get up to the breakout session. Again, that will be on Genie B on the second floor in about 10 minutes. We'll see everybody up there.