SignalFx | Interview with its CEO & Co-Founder – Karthik Rau
In San Mateo (CA), we meet CEO and Co-Founder of SignalFx, Karthik Rau. Karthik talks about his story how he came up with the idea and founded SignalFx, how the current business model works, as well as he provides some advice for young entrepreneurs.
Martin: Hi, today we are in San Mateo in the SignalFx office. Hi, Karthik. Who are you and what do you do?
Karthik: Hi, Martin. Thanks for coming to visit us at our office.
Karthik: I am a CEO and a co-founder at SignalFx.
Martin: When did you start this company and what made you start a company at all?
Karthik: We started a company almost three years ago in January 2013. And my co-founder had been at Facebook for several years and he built monitoring systems at Facebook which was very state of the art different approach to solving monitoring compared to the monitoring solutions that had been on the market for twenty or thirty years. So he and I got together and spent some time looking at the broader market, the trends that were affecting the market to move towards cloud and software-as-a-service and felt that kind of problems that Phil had had at Facebook were the kinds of problems that everyone was beginning to see and without doubt an opportunity to start a company around it and so we decided to start SignalFx.
Martin: Karthik, what is your background and how did it prepare you for starting a company?
Karthik: I’ve been at startups pretty much my entire career, at least, when I started a company that’s been startups. Right out of college I joined the company called Loudcloud which was in the cloud computing space before cloud computing really existed as a market. In fact, we had the name cloud in our name Loudcloud back in 1990/2000. I was there for a few years. And then I went to VMware and I was there for a long time. I started when the company was still relatively small as a startup. I got to grow up with the company and see it grow through various stages of growth and by the time I left it was almost a ten thousand-person company. So that was a real highlight, being able to be a part of a company from the early days and see it through to become a fairly substantial and significant player in the industry.
Then, I went to another startup called Delphix for a few years but I always wanted to start my own company. When Phil and I started to talk about building SignalFx everything just seemed to click and it seemed like the perfect opportunity to go to realize my dreams of being a founder and build a company of our own and so Phil and I decided to start SignalFx.
Martin: Can you shine some light on the first 12 months of starting the business and the operations? How was it like?
Karthik: It’s actually pretty boring for the most part but our model is a little bit different. I mean, every model is unique here in Silicon Valley but Phil and I were able to raise money right away so that the two of us had.. I think, because we’ve both been in the Valley for a long time and Phil had a lot of credibility around the technology because he built a similar kind of solution at Facebook, we were able to raise money right away before we hired our team. And so we had the luxury of taking a little bit of time to build the right product and do it thoughtfully.
So from the time we started the company until we launched the company it was almost a full two years. So the first year, really a lot of it was, you know, laying the foundation, determining what the technology and architecture was going to look like; recruiting our initial group of engineers, building out the core technology and getting to a point where we could begin to go and recruit beta customers. But the first year, it was pretty much all product development. Beyond at the very beginning there is the fundraising process, and then once that was done it was pretty much just building.
Martin: And when did you start talking to initial customers?
Karthik: Well, we had beta customers – I think we started our first beta about a year and a half after we started the company, roughly about a year and a half after beginning. And I guess the question is how did we find these customers are or what the -?
Martin: No. Just when did you start looking for them so one and a half year – ?
Karthik: Yes. Well, we wanted our product to be far enough along that we could actually deliver value to customers but we didn’t want to wait so long that we would go too far down a path that wasn’t necessarily appropriate for the marketplace. That is a very fine line to walk. We felt we hit that point about a year and a half and where we had enough of the core technology that could actually solve problems for customers who had a big enough pain point in the areas that we were addressing. So we found some of these early customers who were willing to invest the time in a young company and be design partners to us and work with us on our beta. So, yes, it was about a year and a half after we founded it.
Martin: What was the customer feedback like? Did they say wow this is really awesome or did they say you are on a good track but actually we would like to change this and this in the feature in another way that could solve our pain points more efficiently?
Karthik: I think generally the overall feedback was very positive. Since the very first customer has signed up, they loved the product and loved this base they were in. There is always feedback and the feedback tends to be in the places that you didn’t expect, right? So little things that you missed because you think about the product the way that you would use it and then you go to a different organization that has a different model and maybe a different process and or a different way of doing things and you have to sort of adjust it a little bit to help solve their problems in ways that makes sense with their organizations. So those were the kinds of things that you have to build small features to help our product integrate better into a customer’s environment and address them, scenarios that maybe we haven’t thought about.
But the core concept, the core product, the core value proposition we’ve never wavered from. almost three years ago we thought the value would be, continues to be where most people are getting value today and what people are excited about today. It is on the perimeter, there are a number of things that, as you engage with customers you realize that we really need to add these things because it’ll make it much easier for customers to use and deploy and get the value out of our products if we have these additional components.
BUSINESS MODEL OF SIGNALFX
Martin: Karthik, can you, please, elaborate a little bit on the value propositions with what type of pitch are you currently targeting your customers?
Karthik: SignalFX is a monitoring service for monitoring application. So we are now not focusing on monitoring traditional enterprise apps like a SAP, Oracle. There’s plenty of companies that have been doing that for the past thirty or forty years.
But what we do as your building the next generation of new applications. If you’re moving to a cloud architecture or even if you’re running your own data center but you’re building a modern application that has micro services or open source or it’s just distributed in nature it is a very different challenge monitoring these modern apps versus traditional apps. And the big differences are:
- Number one, these applications tend to be updated more frequently. Instead of doing one or two updates a year, you have updates that happen once a week or once every couple of weeks.
- Number two, they tend to be more complex applications where it’s not just one app running – one monolithic app running on one machine. It tends to be service oriented so you can have multiple services, all communicating with one another. So if something goes wrong you have to understand and pinpoint did it go wrong in service A or service B and how does that affect your entire chain of services.
- And then the third thing that is fundamentally different is that apps tend to be distributed. So you have because the VMs and containers are so lightweight and so easy to spend up in cloud environment you could have applications that have dozens, hundreds, thousands, tens of thousands or hundreds of thousands of components.
And so in that world if you have to monitor large complex rapidly changing environment our belief is that the only way you can do that is through analytics. And you need to collect a bunch of information across all these different systems, perform real time analytics on it so that you can more practically identify trends and patterns as they emerge in your environment. And that’s really our specialty. SignalFx is – we solve the monitoring problem but we do it through analytics. And so for our customers the benefit that they get is number one they get real-time complete visibility of everything that’s happening to their applications. So keep in mind most applications these days are living breathing organisms and more and more they are critical to business. It’s not like in the old days you have some back-office applications that is unavailable for a little while, it’s a nuisance to people. But today these applications are your business. And if you have problems with that it materially impacts your business. You want to understand everything that’s happening as it’s happening.
So we provide that visibility, we provide very high quality alerting. And so instead of getting alerts every time some random component is failing you get alerts when some meaningful trends are happening like significant subset of users in a particular region connecting with a particular device are experiencing larger than normally lengthencies. That is something that you care about. Those are the sorts of things that we help our customers really understand very, very, very quickly and very effectively.
Martin: If I as a customer of yours, for example, have a very distributed kind of API system – how do I set up your SignalFx service?
Karthik: There are lots of different ways of doing it. We have technologies that we provide out of the box.
We have an open source agent that a lot of our customers use. It’s one-line demand to get it installed and it collects all of the infrastructure metrics; it collects a bunch of application metrics if you’re running open source components. If you’re running MySQL, Postgresql, Elasticsearch, Kafka, Cassandra, these common open source, the platform will collect metrics from right out of the box and give you a comprehensive analytics and dashboards around it so you can understand what’s happening to those.
For your custom apps or for your APIs you would probably have some of those metrics put into log files and so we have log scrapers that can take those metrics, count them and send them into SignalFx or you can just send metrics directly into SignalFx from your applications. Every time an API is called, for example, you can send a metric into SignalFx that measures how long it took to process that request must including the number of dimensions around it, like which customer made that request, what region they are coming from. And then once all that data is in SignalFx you can start to slice and dice it in meaningful ways. As events are happening in your environment you can see it in SignalFx.
Martin: Who do your target customers? And how do you basically acquire them?
Karthik: Our target customers – it’s basically anyone building an application who could use SignalFx. And there is a lot of application development happening in the technology community obviously. And so everyone here is a potential target for us. But more and more we see traditional enterprises also investing a lot in software development because… As our investors, Marc Andreessen Horowitz likes to say “Software is eating the world and every company is becoming a software company” and so as they’re investing in new software they also need a system like SignalFx to be able to observe as a software is out there in the wild, the operational characteristics of a software, performance metrics, to understand how the applications are scaling – all of these things that they need to understand they could use SignalFx to do it.
The second part of your question – how do we acquire the customers? We are SaaS service, we do all the traditional marketing that you would expect business to do we.
We have demand generation programs, we have a more concentrated community of buyers today. It tends to be operations teams. Some people also call them development teams. There are conferences where they tend to cluster and so we are present at those conferences. We also have a tremendous amount of word-of-moth with our existing customers. We have very happy customers and they tell their peers and then they come in and they find that since we’ve got a lot of inbound people finding us, hearing about us, coming to our website, signing up for a trial.
We have a free trial. It’s a very lightweight process. Someone can come and sign up, then immediately get value from the product and get a sense of the value they can get from us. So it’s a very low friction, very easy way to get started. It’s not a heavyweight sales engagement like classic enterprise software.
Martin: Are you also using distributional partners like consulting companies?
Karthik: As a young company the most important thing is to understand your own model very well first, I believe. So we’ve only been selling for six months. We launched in March. We’ve built our sales scheme. It’s an exceptional sales scheme. We are really understanding all the nuances of our go-to-market motion ourselves but, absolutely, I think over time there’s a lot of opportunities to work with distribution partners and services partners. But right now our focus is on understanding are own model very, very well before we start engaging with partners.
Martin: It makes sense. Karthik, you said that you are basically is a SaaS model. What are the price drivers in your model?
Karthik: The way our product is priced is based on the amount of data that customers send us. And we’ve done a lot of things to make it very fair and easy for customers to control what data they send into SignalFx. And the reason we do it this way, in the most traditional monitoring solutions prices are based on the number of servers that you are running. Unfortunately, that doesn’t work well in the modern cloud environment because first of all these environments tend to be elastic. We talk to so many customers that could double or triple the number of VMs in their applications on a daily basis. They’ll scale up and they’ll scale down, pick hours will spike up and then they’ll come down. So host count is a lot more fluid than it used to be. It’s not quite as predictable as a metric of usage as it used to be. And secondly, not all hosts are graded equal anymore because you have some applications that you really, really care about and you want to collect a ton of data around those and other applications that are not critical and you don’t collect as much data. And so it’s not very fair to price the same for both of those.
So what our customers do is, they can decide whatever data they want. Ultimately the primary cost of the solution like this is data storage, we just charge them on the number of data points that they send in, the data rate. The more data they send in, our storage class are higher. We have to retain that for our customers, and we tell them okay, there’s a higher cost, the higher price. The less you send in, the less we have to store, the cheaper it is for the customer. It’s a very fair model for customers; they can decide what they send, they can decide how they want to send it and they can download them down at a moment’s notice.
Martin: Karthik, you said that you are basically doing analytics based on the apps pinpointing where there might be an issue. Are you also doing predictive analytics for seeing maybe some patterns between your customers and seeing, there might be an issue for that kind of apps and there is some overlap in the usage of similar apps between some of those customers. I see an issue rising at one customer therefore I may be predictive, tell another customer there might be an issue coming around – here is how you should solve it.
Karthik: We don’t do that because of number of reasons.
Number one, our customers are very sensitive about data privacy, right? So we even anonymize, we don’t do that – there is no pattern matching between different customers. It’s completely isolated; there’s no comparison of one customer to another.
Secondly, I think one of the things that’s fundamentally different is that every application is unique. And we have a massive fragmentation in a number of applications today compared to twenty years ago. Twenty years ago, everyone was running package software, and so you could look at Oracle trends across 10,000 customers. But today, Salesforce is very different from Work Day which is very different from Facebook, very different from Google and they are all unique applications that have their unique workload characteristics and unique hardware configurations. And so it doesn’t make as much sense anymore to do those patterns across customers. There could be some benchmarking sorts of things down the road but at this point we’re focusing more on giving the tools to the customers, the exploratory tools so they can richly understand their own data and look at the patterns in their own data and draw conclusions from that rather than trying to solve problems from looking at patterns across customers.
Martin: Karthik, what have been the major obstacles over the last three years and how did you overcome them?
Karthik: Well, I think as with any enterprise business one of biggest question marks is what your market model is going to look like. I think that when we started with the company we had a lot of conviction that there was a huge need for a product like what we built. And I think that’s been validated over and over again, the more we talked to customers and the more we are around in the marketplace.
I think the big question mark in the early days was what our go to market motion looks like; do we sell more to mid market accounts, enterprise accounts, where is the interest going to be? In the early days you have to spend a lot of time experimenting with different accounts. For us, it’s turned out that there are all sorts of segments. We’ve got some great large customers that are traditional enterprises in large web skill organizations where they’re essentially replacing their entire monitoring systems with SignalFx. We have midsize customers that move a little bit more quickly. And the deal sizes are not quite as big but you know they want to have more of that transaction motion.
And so we’ve seen both models and then I think figuring out how to structure our team so that we can best equip both of those segments to be successful, that has been a kind of our biggest learning over the past year. And I wouldn’t say it’s a challenge but I would say that that was a big question mark when we went into our launch – what kind of go to market motion do we need to build for these different segments? Should we focuse more on one or another or gain the opportunity in both? And we’ve certainly seen the customers across the entire spectrum and so designing the go to market organization to support all of those customers has been probably the biggest area of investment for us as a company in the last 6 months.
Martin: And what type of experience did you do in order to validate your go to market strategy?
Karthik: With sales it was very easy. You see your customers signing up for your service. Are they willing to pay you? And so the experiments are – I think you experiment with messaging and you see which message has worked. In my experience, one of the things that we did very well is we hired some exceptional people in field who are very thoughtful, listen well, very intelligent. They have almost qualities like stem cells; they can adapt and become whatever model we need them to become and it was an incredible luxury to have people like that in the early days because we were able to really listen to customers, observe the patterns that were coming in and be able to figure out, ok, well, this is how we need to design our go to market motion versus assuming that it was going to be one way or another. I think we’re able to adapt very quickly based on what was on the market.
Martin: And how did you find those generalists sales people?
Karthik: We interviewed a lot of people and I think we were just fortunate that we found some really high quality individuals who were willing to come and work with us.
ADVICE TO ENTREPRENEURS FROM KARTHIK RAU
Martin: Karhtik, this is your first company that you’ve started for yourself. What type of advice would you share with other people thinking about starting a company?
Karthik: Lots of things. You can spend a lot of time talking about this. I think what most of entrepreneurs would say is number one, you have to be very passionate about the idea because it is like a marriage – you are in it forever. You need to be excited about the team that you’re starting working with; the co-founder chemistry has to be very, very strong; you have to really love what you’re doing whether that’s the product, the technology that you are building, or the team that you are building. So, I think that’s absolutely critical.
I think another thing that I would suggest to everyone is don’t forget to enjoy the journey because I think a lot of people get so focused on – they are stressed about the outcome. That’s natural. You are always trying to create the best organization that you can, but it’s a beautiful journey. I think for me Mondays are my best days of the week. I get to come in and work with a fantastic group of people. Phil and I get to pick who we want to work with every day. We get influence it in a really meaningful way, the culture, the values that we care about. And that’s a really beautiful thing. I think you have to stop and smell roses sometimes. That’s one thing I think a lot of entrepreneurs forget to do.
I think the third thing I would say is that every company is like a snowflake. It’s all unique. Every company is absolutely unique. And people tend to pattern match and you will meet a lot of really smart people who have had great experiences and been very successful who will try to prescribe their approach to your company and everything changes. There is always some dimension that’s changed. It could be the exact same technology and space from ten years ago versus today where markets are fundamentally different today versus 10 years ago. Your personalities are going to be different from other personalities, your founding team and culture everything is going to be unique so you really have to understand and introspect very well who are you, what do you do very well, what are your strengths, what are your weaknesses, what does your market truly look like, what’s different about the market today versus five years ago, ten years ago and come up with a unique strategy because you are unique, your situations are unique. And you can get incredible input and advice from people but all of that input and advice comes with a bias and if you understand that bias you can then tease out the bias and just try to understand which of these things should be applicable to my world and which are not. But oftentimes it means you have to make a decision that is a lonely decision because it’s based only on your company and what you have to do but it’s very, very important that you make that decision on what’s best for your company and not based on what’s other people telling you. I think that’s a very important thing to do when starting a company.
And then I think the last thing I would suggest is from a personality point of view – I think you have to really, really enjoy the wins a lot more than you get down by the losses. Because every company has its ups and downs and as you go through the life cycle of a company early on by its very nature if you’re in the market that’s a new market that people haven’t figured out as the big market yet you’re going to have more people not understanding what you do very well understands what you do then someone else who is bigger and has more capacity to fulfill the demand with the market. So your timing has to be right from the beginning. You grow as the market is growing and that necessarily means that in the early days you’re going to have fewer wins but you have to really relish the wins. That is everything I would suggest to people.
Martin: You said that every Monday you are very happy to come to the office and work with great people and shaping the company and culture. What do you do to shape the culture of the company?
Karthik: I think it starts first of all with who you hire. Culture, it’s not something that you write up on a wall with a few bullet points. It has to be embedded in everything you do, every decision you make, every person you hire, what behaviors you embrace and reinforce and which behaviors you try to eliminate.
I think it starts with the people you hire so you have to be very prescriptive about what are the cultural attributes that you care about. When we interview people we always have at least one person do a fit screen culture screen. I interview everyone that comes into the company and I’m always the final culture screen. And we’re very prescriptive about what are the cultural attributes that we look for and that is very, very important. We’ve turned away a lot of people that are incredibly talented individuals but we felt that they didn’t map to the kind of company and the culture that we wanted to build. So for us there are things like we really care about intellectual curiosity and someone, for example, like engineers that ask questions about the business and our business model are candidates that we really like. If there’s a very talented engineer that shows no interest in the business, it really sways things to the other. Because we want everyone at our company to operate as business owners and we screen for that from the very beginning. And similarly on the sale side, when we are interviewing sales people we like them to ask and want to understand product strategy. And in that way we get everyone who when they are at the company, they are custodians of the business and they may be wearing a particular hat to be a salesperson or an engineer for their day-to-day jobs, but if they see something that they question there are not afraid to ask a question and then collectively as a team will make better decisions because we’ve got a variety of perspectives.
So that’s one independent thinking as another. Because as a start-up you have to take risks and you have to go against the grain sometimes and you have to make non obvious conclusions and so we really try especially in leadership positions to find people who are creative and will connect the dots in ways that other people don’t. If we learn interesting things in an interview and we see different perspectives that we haven’t seen before that’s positive for us.
There are all of sorts of these things that we try to do to screen for culture. So who you hire, the behaviors that you enforce, who you promote into management positions is one of the most important factors in building a culture, because I never would run a company we will see these are the behaviors that are rewarded. The behavior of managers is also very significant because that’s again what people look at us – this is what the company values. You have to be very careful about who you promote, what behaviors they’re displaying to the rest of the company, how you allocate different responsibilities to people, who you get what projects. Again it reflects to the rest of the company what qualities do we admire and respect. If we have a really big problem that we need to solve and it is very important for the company to solve it and we give these three people the role of solving that problem and it tells the company these are the three that were trusting to solve this big problem and they are obviously the skills that they have that the company really appreciates and respects.
This isn’t just about technical skills. It’s technical skills, it’s interpersonal skills, it’s just all of these things become your culture. And I think it really starts with the people who you hire, who you promote, who you give certain roles too. And then myself and Phil, my co-founder, how we act every day is I think people look to us too, and the things that we value and how we treat people and treat partners and look at the opportunities. I think this is something that people look at as an example. There are so many different things that are involved in building a culture but I think the key is that you have to keep reinforcing certain tendency and do that across the entire company.
Martin: Karthik, thank you so much for sharing your knowledge. It was really the pleasure.
Karthik: Thank you, Martin, very nice to meet you.
Martin: Likewise. And the next time when you are starting a company you should really focus also in the beginning on the company’s culture that you are hiring the right people not only based on the skills but also based on the attitude, because you want to minimize the asshole factor, for example, what do they work and be very curious about what they are working. Thank you so much.
“A leader is one who sees more than others see, who sees farther than others see, and who sees …