Celonis was founded in 2011 by three students who didn’t know they wanted to start a company, but fell in the love with a school project.
Today, Celonis is a Forbes Cloud 100 company, and the leader in a very interesting category, enterprise performance acceleration software, leveraging a company’s data exhaust to understand which processes work and which need to be be optimized, through process mining technology.
It’s also a unicorn startup with $367 million raised to date, most recently at a $2.5 billion valuation from investors such as Accel, 83North, our friends at Arena Holdings (who kindly introduced us to Alex) and Qualtrics founder Ryan Smith, who spoke at Data Driven NYC a few years ago, and then famously went on to sell his company, Qualtrics, to SAP for $8 billion.
We had a really fun chat at our most recent Data Driven NYC, with Celonis co-CEO Alexander Rinke:
- How Celonis started as university consulting project when the founders were 21
- How Alex waited several hours outside the VIP area of a tech event, until he was able to talk to the founder of SAP, which resulted in a transformative partnership
- How the company was bootstrapped for 5 years before taking any VC money
- What it takes for a startup to successfully expand internationally
- What is process mining software and how does it work
- Go to market strategies – horizontal vs vertical
- And a lot more
Here’s the video, and below is a full transcript of the chat:
FULL TRANSCRIPT (lightly edited for brevity and clarity)
[Matt Turck] I’d love to start this conversation with the Celonis story. Maybe let’s rewind to the very beginning, 2011, Munich, Germany. What happened?
[Alexander Rinke] Yes I remember I was just an undergrad, and I was 21 years old, and we did a university project for a local company here in Munich with the goal of optimizing their IT service processes, right? Every larger company has an IT help desk to help employees with issues. They asked why it wasn’t working very well. It took too long to get back to employees when they had issues. It was a slow process, inefficient process, and they had probably around a hundred people working on this process. This was this kind of project where students get an opportunity to be involved in a real business challenge and learn something about business. We were all technical, probably like many people on this call, like all technical in technical fields. My major was math and my co-founders were computer scientists and we go into this and most people don’t really take it seriously. They check the boxes and then move on. We really fell in love with this company and this project, and we wanted to help them out. So what we did is we did a lot of workshops. This was well pre-COVID, so you could still meet people in person! So we would do a lot of these meetings in conference rooms, and we would basically interview the employees about what’s not working, like a consultant, it was really like consulting, like a management consulting project.
At some point, we just noticed, this is so inefficient, right? We talked to all these people. Everybody tells us something different. You never know what’s true. You get like five reports out of their reporting system. We don’t really get what they mean. They fake the KPIs. It’s just such an inefficient way to make an assessment of this organization, the problem. We had learned at university about emerging technology in data science called process mining, which takes logs from IT systems and is able to reconstruct how the processes in an organization that uses this IT system actually are performed.
So we had this idea to build an X-ray for a company. So, we thought, “Okay, let’s build this tool that can X-ray their ticketing system, like a [xyz] and they use that and then all the other systems they use like a telephone system and other systems to get a visual representation of how this process actually works and where things fall down, where the bottlenecks are, where all the friction points are. So we were going to build an X-ray, like a digital X-ray. And we did. We built the first prototype which was really basic. It was like this thing where basically the UI was like a SQL window trying to SQL code and then you would get a PNG file of the process. It was very basic, but what it did was like an X-ray system for business.
That’s interesting. So you started, and then we’ll talk about this in a minute, because you didn’t raise money initially, but when you started, you actually built a product as opposed to delivering services or was that a mix of services and product?
Yes. So actually we didn’t raise money for half a decade. We raised $367 million since. So the company has been around for nine years and the first five years we didn’t raise a cent. And it was always a product business from the very start. But the initial project, which led to starting the product was the services kind of thing, but it was more like a university project.
How did you do it? Were you able to through this university project, find customers early…?
Here’s what happened. The CIO of this company saw the prototype that we built, and he’s like, “I want to become your customer.” And like we were “we are three students. We don’t even have a company. We just put this prototype,” and he’s like, “Yeah, I want to be your customer.” And then we said, “Okay, now we’ve got to start a company.” So we started a company. We had no idea how to run a company or build a company like zero, right? If you had asked us two years before, a year before, three months before, “Would you ever start a company?” The answer would’ve probably been no. We started a company and then something incredible happened. Like they really improved, like within three months, they got their resolution time down for tickets, right? That’s a very important metric to measure, down from five days to solving over 80 percent of their tickets in zero… Like in the same day, without firing or hiring a single employee, just through working more efficiently, based on this data that we would provide. They got tremendous leaps in productivity and in customer satisfaction, et cetera.
The CIO became our first customer. And then he called up a bunch of CIOs that he was friends with. And we signed up the first five or six customers that way. Then we left university and started Celonis. The original name of the company was actually Simulitix which was a terrible name, but then we renamed it Celonis. And yeah, we got started and I can talk about how we signed up more customers and things like that, but that was the initial. From the start I think we were very driven by the impact that this idea that we stumbled on, what extraordinary impact…
The other unusual thing that I noticed as I was going through the history of the company was that you signed an important partnership with SAP a few years in, reasonably quickly for a startup. I’m just curious how that came about. The reason why I ask you that is, it’s a question that comes up so often in startup conversations. Like, “Oh, if only I could sign Microsoft or Amazon as a partner, I could accelerate and then there’s always like some debates on the pros and cons. I’m curious how that came about.
There’s actually a lot of people that ask me, “Have you ever worked for SAP or something like that?” Before we started, Celonis, even when we started Celonis, the day we started the company, I didn’t know what SAP was. I didn’t know the company existed. I didn’t know what the system was. I had no idea, right? But we found out pretty quickly that if you optimize business processes, I didn’t know, something like 80 percent of the world’s transactions still go through SAP systems. Like it’s just crazy. They have a crazy share of the trajections in businesses. So it was pretty clear that SAP would be a core system of record, that we would get our data from and tie into and help optimize the user traffic, et cetera. So that was very clear.
And we had no idea how to build a partnership with SAP. So, and then I did some research on the company, and I found out that the founder of SAP and the chairman is a guy called, Hasso Plattner, which some of you might have heard about. And then the interesting thing is I grew up in Berlin. I didn’t grow up in Munich. I started the company in Munich, but I grew up in Berlin and Hasso Plattner lives in Berlin, like in Potsdam, which is very close to Berlin. And he started his Institute, which is probably 10 minutes drive from where I grew up. I saw that he would give a speech, around his Institute. So I called up my mother and said, “Hey, I’m coming home on a visit.” She would say, “Oh great. You haven’t been here for a while.” And I said, “Yeah, I’m not going to stay long, but I got to talk to Hasso Plattner.”
I went and just went to the speech, with no intention to listen to the speech. It was interesting, but that wasn’t my goal. But my goal was to talk to him. And I had a few slides printed out on Celonis. We’d signed up the first customers and show him the product, what we could do with SAP data. And he literally gave a speech and then left to go to a VIP area that was guarded by security. So I couldn’t speak to him. So I waited for probably around three hours in front of the VIP area until he literally had a stone in the shoe and had to leave the area to go outside and I said, “Hey, Hasso you really got to see this.” And he listened. He gave me five minutes of his time and I showed him. He just instantly got it because, I mean, if you’ve built the main transactional engine for business processes, you understand what Celonis is pretty quickly.
And then he said, “Look, I like it. We just came out with our first official memory database HANA, which could be a really great platform for you guys to test out.” And I said, “We already did. It’s really exciting.” And we had a conversation and then he introduced me to the then CTO. Then we didn’t actually get to speak to the CTO, the CTO introduced another person. But the introduction came from Hasso, which means a lot. Then we forged the partnership with SAP and it’s a great partnership still. We are the biggest product that they sell outside of their own portfolio. They said the biggest, third party product in the cloud. I think it’s been one of the fastest growing ISV partnerships in the whole growth company and startup domain. It’s been really exciting.
It took us almost three years from this point to actually start the reseller partnership. We went through various other levels of partnership before that. One advice I would give entrepreneurs, it’s like a partnership. Some entrepreneurs treat a big tech partnership as this silver bullet. It’s going to solve their go to market problem. That’s not going to happen, right? It’s going to be a very long journey and going to take a long time, require a lot of aligned incentives. It’s not that straightforward, but it worked well for us and SAP.
Maybe just one last topic in this entrepreneurial journey, maybe talk a little bit about the international expansion. So how do you go from a Munich-based company to a global company and I think you now are in Germany today, I think you live in New York, if I understood correctly. What did that take and some of the lessons learned along the way?
Happy to talk about it. I mean, I think the first thing in every company that you do, and then you’d be super serious about in the early stages, is putting really strong product market fit, right? I think people sometimes underestimate it. If I talk to a VC, I always say, I have product market fit, of course, you guys would hear it all the time and that’s fine, you need to raise money. But as an entrepreneur, you want to make sure that you start really scaling once you actually will product market fit, right? Meaning you know what to sell, who to sell it to, they’re willing to pay for it, they’re willing to use it, they’re willing to renew and you really figure that out.
And once we figured that out, we said, “Okay, it took us a few years to do that. Now we really want to make this a bigger, much bigger company.” It was clear too and I think the first thing we noticed is that, we have this principle at Celonis we call FISA. Focus, Invent, Simplify, Act and it’s really about focusing and simplifying things. So the first thing we learned about international expansion is you’ve got to be extremely focused on how many markets you go into, how quickly, right. Most of these markets, especially when it has to do with data relatively large, so it’s better to reach penetration in some core markets than to try to spread yourself too thin all across the world. It’s not going to matter. So for us it was very clear that the big bet in terms of international expansion will be the US. It’s by far the largest enterprise software market in the world.
I think that companies that have started in the US have the luxury to wait a little bit until they think about international expansion. Companies that start in Europe often have to do it earlier to become relevant, because you really can’t be relevant without being very strong in the US so we said, “Okay, the US is going to be the market that we’re going to tackle,” and that was in 2016. That’s also where we raised our first venture capital around because we were completely bootstrapped. We just didn’t know how much investment it would take also network and expertise to get into the US and then, I decided to take it on because it was such an incredible experience again. It was almost like starting the company again, with more experience, with obviously more product behind and more customer validation, et cetera.
But I think what we learned is, especially when you move into the US, but I think it’s pretty similar for American companies when they move to Europe, is that’s a very serious move, meaning you need to really, do it very well. You adjust your messaging to fit the culture. One of the things I learned, like we were super successful, the US took off like this. It just tripled again, and it’s growing extremely fast and it’s also our biggest market now. So, it’s been an incredible… I mean, but what we noticed is, you really need to make sure you take it very seriously. You are very focused. And then, we started in the US again, I always say, we came on the Mayflower, with two guys, me and another guy, we went over to the US to start the business there and you do all the things that you do when you start the company originally, except for building a product. You hire a team, get the first customers, you market, you build the company.
I think that probably the second key lesson I’ve learned when starting the company for the first time in 2011 is you just have to assemble an extraordinary team. I think that that’s another one of our core values and beliefs is that the best team wins. And that’s just so true. Every time I’ve compromised on that, I’ve regretted it. Every time I’ve stayed true to that value, it was just extraordinary. So that’s probably one advice I would give you. The other advice, as you go global, you have to make sure that you create a glue between the culture and you really create one culture. So you also really have to think about how people work in the individual countries. What are the cultural backgrounds, et cetera, like Americans are different from Germans and so you’ve got to really make that dynamic work. The biggest thing is to hire really good people.
Switching gears to the product itself, so at a pretty granular level, how does that work? You basically have a platform that has little tentacles in every repository and tracks the data, or how do you get the data into the system so you can mine the processes as you defined earlier?
Yes. Perfect. Great question. So, basically there’s different types of data sources, right? So one is the transactional systems, SAP, Oracle, Salesforce, et cetera. And you typically connect either to the database or to an API, right? So if we connect to SAP or if we connect to Oracle, we pull thousands of tables from these systems. And so also, we have over a hundred pre-built connectors to transaction systems. So we build standards around this, so it’s easier for customers to get started.
Then there is task data, which happens on a user’s desktop and is not actually recording in a transaction system that will be opening an email or looking at an Excel sheet. We can also track that data through basically little tentacles that we have into the desktops, anonymize the data, get rid of all the private data automatically and then pull it into our system to compliment this transactional data with task data. And then there’s some extra data that we might pull from a Nielsen or like an external data source. You can pull that in through API. Yeah, benchmarking data, but between our customers, so that’s easy because the data is already in our cloud. It’s a fully cloud based platform. But, I would say the data extraction process depends on the source of the data. There’s multiple sources of data and then we’ve built a lot of standards around that.
Presumably that’s pretty noisy data. I mean, that’s very different formats, different sources. You have to clean everything up and to ingest-
Yes but the interesting thing about Celonis is that’s what the process mining algorithms do. So we don’t do any manual data cleaning. We actually want to see all the noise because we want to find out where the noise is in the process, right? The system is good at finding what’s just an automated process and where’s the human actually wasting time and you can do that automatically. But we don’t have a big data harmonization cleaning effort. I will give you an example. Like we have a customer, Siemens, one of the biggest deployments and it’s just a huge company. They started early, they pull in 16 billion records a day. So there’s a lot of data that goes through.
Once you’ve gathered the data, then what is the next step? And maybe as I think through this, maybe for clarity, spend a minute defining process mining – maybe give an example of a process that you would mine and improve and then we’ll go back into how that actually works.
Imagine you’re shopping online, okay? So you buy something on Amazon. There’s different steps to that process. You start somewhere on the website. You go buy something. You pay. It goes into the warehouse, it gets shipped. Then you get a bill, it gets billed. So that will be a process that would be something we define as a process. There’s multiple data sources. There could be data in the website or there could be data in the transactional system, the various transaction systems Amazon uses. And it could be data on the desktops of Amazon employees. If there’s an issue with your order, obviously there’s a process with a high degree of automation, but also a high number of transactions. Or imagine you’re applying for a credit card. That will be another process that could run through multiple systems.
So what the system would do automatically is find out how does this process work? Meaning what is the sequence of flow? What is the duration of flow? What is the variation of flow in the most frequent pathway this process takes? So if everything goes well, we call this the happy path. And then what are all the deviations? What are root causes for these deviations? Every time you order something from Tampa, there’s a delay or everything we ordered from this product group there’s a delay. There’s a huge correlation between product group and delay. And then, for example, there’s a huge correlation between individual products from a product group and the amount of returns you have, so that you can say, “Okay, maybe the product description isn’t accurate because people send it back all the time.” So essentially you have a process. It could be any everyday process and it goes through various data sources.
We are very good at collecting this data. And then in very noisy and high volumes of data to get an X-ray that is very easy to read and finds friction points automatically to detect where the inefficiencies are and then we actually go a step further and trigger automated improvements and actions from that.
How does that part work?
Essentially, if we notice patterns and we add some sort of domain knowledge around certain scenarios, like e-commerce, like accounting, et cetera. We can right away automate an action. Our users can also define that based on what we see in the data. So the data is synced in real time and then we trigger actions automatically based on what happens. So, for example, if you see that very frequently for some product, there are so many returns and it didn’t seem to be the price or more the description of the product. You could automatically trigger the process for the product description to get updated as an example, right?
How does that manifest? Whoever in the enterprise is in charge of writing the product description would get-
Would get an alert with a recommendation, say, “Hey, you take care of this.” There are some things that you can fully automate. So I’ll give you an example. With COVID, every company was looking to preserve cash. Every big company was like, “Okay, suddenly we have dips in revenue. We have uncertainty. We need to keep cash as tight as possible.” So we have an app that looks through all of your invoices and figures out which invoices you can pay later than you pay them today, right? Because you don’t use the full term of the contract, et cetera. And when we see that invoice is about to get paid early in the system, we just block it and delay it. And there’s something like we have a customer that based on this payment timing, like telecoms, a big telco, they save $40 million just using that app. There’s a huge value around this. We add domain specificity and we have a partner network creating these apps on top of our platform that address specific use cases.
So ultimately you really disrupt consulting, as an industry? Like instead of sending a bunch of people to look at processes and do PowerPoints to suggest how one should improve a process, you automate part of this, is that fair or not?
Yes. I mean, I think that’s fair. We don’t replace consulting or anything. We disrupt it for sure, but we have thousands of certified consultants from KPMG, from McKinsey, et cetera that use our platform extremely heavy. A consultant can actually help a customer get these insights. A consultant can productize their intellectual property in these apps that we have. So consultants can actually be great partners for us, are great partners for us. But their value proposition shifts from creating a lot of value by generating insight and doing assessments to having that fully automated and even having some of the actions automated to implementing this product, providing guidance to the customer, which areas to look at first, providing industry specificity to the use cases, things like that.
So we should probably open up to questions soon, but, and I’ve a lot of other questions because this is fascinating, but maybe just one topic given the nature of the event, the machine learning part, where does it come in here? Is that for the recommendations that you make?
We use machine learning in various areas for clustering, abstraction, but recommendations, predictions is a big thing. I mean, I look at machine learning as if you have a ton of data, you can have much better predictions, much better classifying, much better models and things that you would accomplish, it could accomplish. Just getting a model with machine learning you can do better. So we use it in tons of different areas.
How do you think about the interplay between business rules and machine learning? I mean, presumably there’s certain areas where you want to predict that, okay, maybe that’s a description that didn’t work, but sometimes you’re like, “No, no, send that invoice faster.” Like this is hundred percent certainty.
Yes. And there’s a use case for both. People sometimes talk about the self-driving enterprise, right? I mean, actually, we can’t even get cars to… Maybe we can, but it takes us years to get cars to drive on their own, right? How long did it take you to get a driver’s license? It just took me a couple of weeks, right? How long would it take you to run Proctor and Gamble? We’re not going to have a self-driving enterprise. You need employees and human input, but you can automate some things in order to augment anything. I think it’s not a question for either or, in the end.
Great. All right Jack, should we open up for some questions from the group?
[Jack Cohen] Okay, awesome. So we’ve got a bunch of great questions from the chat. Actually I’m going to start with a three prong question around data security from Hera. So how do you address or handle data security? Does your solution touch customer data or sensitive PII? And if you do, how do you comply as a global company with data security, privacy regulations?
We started selling to public organizations in Germany, who probably have the biggest data protection standards in the industry, and obviously we have a strong European business, so obviously the GDPR is a huge topic for us. There’s many things that we could explain to the program in half an hour, but we never take any risk on that. There’s many things we anonymized the customer is fully transparent to the customer. There’s solutions where you can keep the data on premise if you don’t want to put into a cloud system. We have Gulf [check, 55:15]cloud solution for the American government. There’s tons of ways to address it. But the important principle is that it’s a super big priority and it always comes first.
That’s great. Another question from Irene more on the team building side. So how do you define extraordinary and maybe, asked differently, is there an example of a time where you compromised on team and regretted it?
So if everybody here who manages a team or wants to build a company, I can just recommend this. Really make sure that you get the right players on the field and then you’ve got to make them work together as a team, which is also a big part of leadership. But it all starts with getting the right players on the field. And I think that I always look for three things. We call them the three Cs. It’s character, commitment, and capability. Then if you check all the three Cs, you’ve got the fourth C, which is Celonis. They are not in random order. They are in purposeful order. Character is the most important thing. If you’re not humble, if you don’t have the right spirit, if you don’t have an ownership spirit, if you don’t have a collaborative spirit, if you’re more about your own success than the team success, you can be the smartest person and the most committed person in the world, this team is not going to succeed. And I’ve learned this with my co-founders, to start, and then with the whole team that we’ve built. Character is just such an important element and especially when you are on an entrepreneurial journey and you want to build something where you really don’t have time for any politics, where you don’t have time for ego and any of those things. So character is the main one.
Second is commitment. You can’t really teach commitment. There’s people that are ambitious, there’s people that are not. And you want people that are committed. And then third is capability. And that’s actually the one that you can make … I mean, obviously you need capable people, but you can sometimes compromise in capability because that’s something people can learn, whereas it’s very hard to teach character or commitment.
I think it’s super important, super, super important. We’ve learned this, as well, for our key priorities and I think that you can never predict the market challenges, as we saw with COVID. The company trajectory, whatever is going to happen, competition, but you own the team that you build to tackle these challenges.
Quick question from Robert. How did your commercial model and pricing begin and evolve, as you scaled?
That’s a very good question. I think that, again, it’s a slightly long conversation, but what’s important is that we always try to tightly price, based on the value the customer receives. We always price the individual components of our product so that it’s much more value than what it costs, but also there’s a clear relationship between price and value. We initially priced the platform based on a process-based model and then went to a user and capacity-based model, which was much easier. We also always wanted to create a company with good fundamentals because that way we can take care of our customers better. And I think pricing is a huge element of creating good fundamentals.
So I think a lot of entrepreneurs, if they sell to the enterprise, underestimate the effort of selling to the enterprise and end up selling their product so cheap so that the sales process is so cumbersome that you can’t really invest into it. So I would always recommend thinking about what’s our sales model and really thinking to align the pricing both with the value of the product, but also the unit economics of your sales model.
One of the questions I have is actually the perfect segue into it, what is the sales model at Celonis? Is it all A-driven with SDRs and BDRs, like a full on enterprise sales force type model?
It’s a full on enterprise sales force. We have introduced digital trials where, for example, you can go on our website … and by the way, everybody should do that … sign up for Celonis Snap for free and just use the product. Then you can go into pay tiers and things like that, but if you are from a big company there’s a chance that somebody might give you a call, but you can test our product. It’s very easy to sort of onboard. Even for COVID, we launched some rapid response apps to help customers out that are free on our Snap platform. So there’s a digital nature to that, but it’s an enterprise and we sell it to the enterprise and it’s an enterprise model. Yep.
How do you think about headquarters in Germany versus headquarters in the US and what were the considerations that went into that decision?
We actually dual headquarters in Germany and the US. I think you can make different decisions. Different things make sense for different companies. The one thing is to be very clear about it. I think that’s sort of a big element. I think that we’ll think about this question completely differently in six months than we did a year ago and than we do today. So I think that, previously, you needed to have an extremely clear location strategy. I think, now, you have to have a strategy for how to deal with locations, in general, and deal with the new reality. We started in Germany, we’re a German company, but the US is a huge market for us and a huge base for us. So we just decided we want to be a global company and have two strong hubs, one in the US, one in Munich. And we are very clear about what’s done where and who’s where, et cetera. You ask me again in six months. Certainly, the last months have been a bit different.
What’s next in the product evolution? What are you most excited about? What’s coming down the pipe for first launch?
So we’re adding a lot of intelligence. I would love to share everything with you. I can’t, but there’s a lot coming and it’s all about accelerating the time to value for our customers. Honestly, that sounds very basic, but for every entrepreneur out there, for everyone building a product, that’s really what you need to think about. There’s so many people that add things to their product roadmap in order to add cool, fancy features to their product. I think the only thing you should think about and you should be extremely FISA about prioritizing your product roadmap should be what accelerates time to value. I mean, it either gets more value in the same time or it gets the same value quicker, or both, to your customers.
[Matt Turck] Since we have a couple minutes left, another question on go to market. How do you think about the interplay between horizontal and vertical? I would assume that, if you serve finance customers or healthcare customers or industrial customers, there’s some specificity to the processes or is there a part of the product that’s industry-specific, one, and I guess, two, is there a vertical-specific aspect to the go to market?
It’s an extremely smart question and I appreciate that. I think the key thing there, I think many companies have the opportunity to verticalize their product. And we do, but now we are much larger. It’s just also, again, a question of FISA [Note: Focus, Invent, Simplify, Act, see above]. It’s harder to sell a platform than a solution. You tell me you have a solution for a problem and, if you can start by solving the horizontal problem that has a big market and penetrate that, do that, and then you verticalize, you get more verticalized, you get bigger. But it doesn’t make sense, if you have 20 salespeople, to have 20 verticals. Right? That’s how we essentially did it. We started very horizontal and then, as we grew, it became a vertical. Of course, we realize, three to four years, it will be very different. Right? Because we evolve. To all the sort of earlier-stage entrepreneurs out there, I would just be very FISA about that question, but I think most platforms have the opportunity to do both, have some horizontal solutions and then grow into more vertical-specific ones. And we certainly do at Celonis.
I mean, I have another hundred other things, literally, and this is fascinating.
I’ll be back, guys.
Such a story. We love the hustle story at the beginning is amazing and what you guys have built and it feels so meaty and important and it will, indeed, unlock millions of dollars on savings. It was very, very impressive and we appreciate your joining the group today and telling us about the story. It’s fantastic. Thank you.
Thank you so much, guys.
Awesome. So I am promoting Nate and depromoting Alex. Thank you, Alex.
Thank you, guys.
Thank you so much.