In Palo Alto (CA), we meet Co-Founder & CEO of ThoughtSpot, Ajeet Singh. Ajeet talks about his story how he came up with the idea and founded ThoughtSpot how the current business model works, as well as he provides some advice for young entrepreneurs.

INTRODUCTION

Martin: Hi, today we’re in Palo Alto in the ThoughtSpot office. Hi, Ajeet. Who are you and what are you doing?

Ajeet: My name is Ajeet Singh and I’m co-founder and CEO of ThoughSpot. We started this company in 2012, I’ve been in the Valley for about seven years, actually eight years now. I came here in 2007 and joined Oracle for a start-up and started a company called Nutanix in 2009 then I started ThoughSpot in 2012.

Martin: How did you come up with that business idea of ThoughtSpot?

Ajeet: The business idea of ThoughtSpot was primarily I would say driven by the experience I had at Oracle and then the first startup where I worked on Aster Data Systems that was in 2007-2008 and I saw that the data infrastructure was becoming very scalable, we call it big data now. People were coming up with more scalable model of storing lots and lots of data in a cost efficient way but how this data was delivered to the end users that was not changing that was still done using the traditional reporting model and that’s very painful because it requires a lot of work and experts you need technical experts to build reports for business users. It’s very slow and inefficient process and that’s where we saw the opportunity.

We saw that in the consumer space there are two billion people using search and the way to find information and we thought what if we could actually provide that user experience for numbers? What if marketing manager would go to a simple search bar and ask for revenue in a particular quarter for a particular campaign they might have done on a product line or whatever they might be interested in and just get the answers on the fly without actually having to go to somebody to say: “Hey, can you brief me a report?” And go back and forth on what the report should be like, changing it becomes very painful and those kinds of things.

We saw there being a huge difference in how people were accessing information at home through Google and Facebook and Linkedin and at work through traditional business intelligence reporting technology and that’s the gap we’re trying to fill.

Martin: How did you start about it? How long did it take for you to build the first iteration of the product? When did you raise money? When did you talk to first customers?

Ajeet: I would say we talked to first prospects even before we started the company. So early 2012 we spent time just learning the market. We realized that this is a big market and there’s a big problem.

So my way of thinking about startup has been driven more by markets. Picking the right market first is the first matter for me has been the most important thing, the idea actually comes later. What is the market and what is the problem that we can solve? There can be various ways of solving it. But before I get down identifying how exactly we want to solve the problem I want to spend time with customers, I want to look at competition, I’m going to look at partners and understand how we might sell a product to them, how people will buy my product in those markets and things like that.

So we spent about six months just studying the market and talking to prospects, talking to partners talking to be people who sell to these companies, just learning. And once we figured that out then we learned about the problem obviously and then we went to the white board and we said, “What would be the most awesome way of solving this problem? The most simple way?” Because technology has become very complex and lot of innovation now is just bringing simplicity to complex way of doing things because complexity means you’re going to be slow and it will be very expensive to do things. Make them simple, lot of people can do those tasks and then do it more efficiently and they can do it easily.

So we came up with this idea that if you apply search to analytics it can actually be 10X to 100X difference in how people operate today. Then we started defining the product, defining our architecture of the product and spent about six months on designing the product. When I say design I’m talking about the technical design, some part of UI design. Because what we’re building is very UX driven but in the back there is a lot of complexity that we have to deal with. So it was important to make sure that all platform is architectured well so when we go to market we don’t have to go iterate and iterate.

Martin: At what point in time did you think about talking to investors?

Ajeet: So this being my second startup. I already had some existing relationships with some investors. As I said my previous company I started 2009 and that company has actually done really well so in the process I had the opportunity to work with Lightspeed Ventures and Khosla Ventures and both of them are great investors great people. For me it’s only been about people because the kinds of things we do for some people it might look crazy. Only thing that is guaranteed is the experience of working with awesome people. So it was basically working with the same sort of folks, making me sure that we are going after the market and a problem that is highly valuable, we spent time on that and raising money itself was I would say not the biggest challenge that we had to overcome.

BUSINESS MODEL OF THOUGHTSPOT

Martin: Ajeet, let’s talk about the business model of ThoughtSpot. So what are basically the customers that you are targeting and especially in terms of the functions that maybe more adoptive to your solutions? And exactly what is the value proposition besides only being simpler to get access to data and analytics that you deliver to those customers?

Ajeet: We currently are focused on mid to large enterprises (Global 2000), that is the way I would categorize my market segments and within these we are building a solution that is designed for business functions (sales, marketing, operations, supply chain, finance, and so forth) but what we’re building is very scalable. So we can go to a large company and potentially they might have 50,000 people that would want to use ThoughtSpot. So this is a solution that can start small with a few hundred people and then on time you can scale to thousands of people.

So we do also get involved with IT departments, so typically a director or VP of Business Intelligence would be someone we will to talk to, or somebody from the business side. And the value proposition for them is imagine you were talking to marketing person, a head of marketing, whois primary responsibility is to reduce customer churn, then they have currently existing way of looking at data of how the churn numbers are over time and how different campaigns might have an effect on churn. We would go in and demonstrate to them with our technology they can get access to their churn data in a much more granular and much more ad hoc way, because they’re launching campaigns and this is actually a real example. Previously it would take them two to three months to find out what the effect of the campaign was on their churn numbers, now they can do that in a few minutes. So launching campaigns and next day as soon as the data starts to come in I can go in and understand how major numbers are changing by geography by product line on those kinds of things and then based on that I can then optimize which campaigns I do more, which campaigns I do less, and things like that.

We now have customers that are across almost all major industry verticals (retail, financial services, telecom, manufacturing, and so on), so we work with a whole range of people.

Martin: From my point of view the major benefit of ThoughtSpot is that you’re making it super easy and accessible for people to access the data in a company. How does a customer of you ingest the data into ThoughtSpot and how do you set a specific user rights to data because maybe only the c-level management would like to have access to financials and some other people should not?

Ajeet: So we’re doing to data / to enterprise data what Google did to newspapers. Long time ago people used to get information from newspapers and it was published by a few people and it was then formatted, it had headlines and there’s a front-page and sports page and so on. It was very rigid in structure because only a few people could publish the newspaper. But with Google now anybody can go in and find information they want and the news they want and things like that.

When we apply the same model to enterprise the user experience model remains the same but other things change. The kind of things that changed I would say number one is the type of data we’re going after. Google and most of search engines that are out there, the data they’re ingesting are documents, there are web pages, radios, videos and things like that. What we’re going after is, we are connecting to the most valuable data assets in large companies. So that data sits in their ERP system or data ware houses or any kind of databases, now we have Hadoop, it’s complicated environment.

So the type of data we have to deal with is very complicated, so we had to build a new kind of search technology that can understand all the complexities in the databases like tables and joints, columns and things like that. And it was done from scratch.

Now when you’ll use, search and analyze this data the performance expectations are also completely changed because in our traditional reporting environment your report comes back in 20-30 seconds and that’s considered fast. But if you’re in the search bar, the user expectation is very different. So our search engine is backed by a very high performance computational engine that can take data from hundreds of tables and billions of rows and do joints on the fly and give results to end users at certain speed.

The third big change is security as you said. People want to get enough data across departments, across different levels and things of that nature.

And last but not the least, I will say the biggest way in which we differ from a classic search engine is accuracy and trustworthiness of the data. Because if I am looking for friends on Facebook or i’m looking for books on Amazon, or I’m looking for coffee shops on Google I’ll get probably 10, 20, 30 results for any given search where I’ll have to pick which of them is relevant for me. There is no guarantee that something is going to give me an answer which I’m interested in, but if I’m giving ThoughtSpot in the hands of a hundred marketing analysts or hundred financial analysts or thousands of people in an organization that might be doing very specific functions, they are used to getting data from reports or Excel sheets. They will not accept guesswork.

So we have to build the technology that does not do any guesswork, it actually uses intelligence that is already there their data and uses that combined with the inputs from the user through the search bar and always gives them one single result. Even though on the front it’s very simple, in the backend we had to handle scale and complexity of data and we have to handle security and governance. And the nice thing is that when we go to our customers with ThoughtSpot the business users are happy because all of them can access data on the fly whenever they want to. IT department is also happy because today IT department is buried under a long list of report requests and they’re always behind. Some of them told us they feel like report monkeys and they build more and more reports as opposed to going to the business and saying what kind of data can I give you so that you can make better business decisions.

With ThougtSpot IT can be in the business of finding new data sources, provisioning new data sources and putting security and governance around them and business users can access the data that they’re supposed to access.

Martin: Ajeet, how did you acquire the first one or two customers? How do you convince them: “Guys, please give me access to the most precious data that you have and I give you some kind of answers to that”?

Ajeet: In the Valley and elsewhere as well there is a model of working with your investors to get access to initial customers. We also do that and it is extremely helpful in making introductions to potential customers. But at least for me the way we like to build the company is it has to be done in an organic way and what I mean by that is we have to find a natural fit between what we are building and the segment of the market that will be interested in.

When I’m going through the process of defining and building the product the exact specific market segment is not that well known and also how I place the product to them and how they will they use it, what’s important and what’s not that also is not known. So you go through this process of finding the product market fit. The product market fit is better if it is done organically by the company itself where you might have… I would say 18 months before we had a product we had an inside sales person who was reaching out to people, calling them and doing demos. About a year before having a product, we had a full sales team and these are expensive sales teams, you can spend anywhere from half a million to a million every year on sales team. But for the kinds of things we are doing it typically takes 12 months to figure out what exactly is the product market fit. Is my product good for small companies or is it good for big companies? If it is good for big companies who are the people that I need to talk to? Who will be more in the buying process and what do they look for? What features are important? All this has to be discovered in the field organically.

If you just depend on your own network which is friends and family, or you depend on investors then you’ll go to people who will try your product because there was an introduction that was made or they are your friends. And that might give you false positive because they would say: “This is awesome, this is great. I love it.” You don’t know if they’re saying that because they love you or they love the product. They can also be false negative because maybe they’re not the right users for the product that you are building the right user is somewhere else.

So I think it has to be a good mix of reaching out to the network as well as organic outreach with a bias I have a personal bias towards organic outreach. And in our case this is what we did, we reached out to people organically, we showed them the product, we got their feedback and where the best fit we went after those. Since we were building enterprise product and the data is very sensitive (that’s what you were hinting at), it is on-premise product so we’re actually going to deploy our product inside the firewalls of a customer so that they don’t have to worry about data security and those kind of things.

Martin: Did provide your first customers just free trial to show the value and then after you’ve shown the value then you said: “Okay, let’s put some revenue numbers behind that”? Or did you start in the beginning: “I would like to put some dollar sign on that”?

Ajeet: Yes, even our beta were paid, our first beta were also paid. We didn’t want to give our product out for free even for trial because it leads to the same situation I was talking about earlier. You end up with people that just want to kick the trials either because you know they have fun keeping trials with new technology or taking it because they’re doing a favor to you and your investors. But if you ask for even $1, asking for $1 is so much more valuable than giving your product away for free. So the goal there is not to make money and fund your business, the goal is to make sure that the problem you’re trying to solve is a real problem.

So if I go to a business and it is let’s say an accounting firm that wants to understand how should I do all my business and which of my customers are the most profitable for me that has to be a real problem so that I know you if the product actually works there is product promise that has to match with the business problem and then value can get created. So it’s very important to make sure that you’re going after real problems and not just trying to to get lucky or have some good accidents along the way and eventually get to success.

Martin: Ajeet, let’s talk about the technology of ThoughtSpot. If I’m a potential client and I’m signing up to ThoughtSpot, how does it work then? How is your machine learning working?

Ajeet: The way we thought about building the product was we want make it simple on the frontend for the end users but we also want to make it simple on the backend for IT to set it up. Because traditional products they’re very clunky for business users to end users we call them humans so we like to call ourselves analytics for humans, average human beings might be experts in sales marketing but not in analytics should be able to access data. So you want to make it simple for the humans on the frontend but on the backend also there is a lot of work that is done to set up BI products. You have to kind of enter the data source obviously, then you have to talk to the business users and say: “Which reports do you want?” And for those reports I have to identify where the data will come from, what kind of data model I’m going to need, if the query is going very slow I might have to build what is known as cubes (it’s basically pre-aggregation).

I would do those computations over night, so next morning you can look at your revenue for each country. But if you wanted to look at revenue by each product line: “Oops, I didn’t do that. That’s a problem with that.” So you have to define all these things in advance.

With ThoughtSpot we leverage lot of memory computing at scale and we have cut down on any sort of pre-computation that is required and every step we tried to make things simple and cut down on any human intervention that might be required. So what we do is we will go to your signing up with ThoughtSpot, we would implement our product, install our product in your data center. That is typically done within a day and we will get you live within 2-4 weeks from day zero. And that’s just to give you a sense. The classic BI deployment takes about six months, so we have shrunk that time of six months down to 2-4 weeks for some of the biggest and most complicated data sets that are around there.

The way we do this is we connect to data sources and we then build sort of a mirror image of the scheme of these data sources. And then our search engine, as I was saying earlier, has been built to understand all kinds of things that in our scheme, it will understand all the tables and joints and lots of meta data that is already present.

A lot of times customers say: “My data is dirty” and it is not going to be ready for this, which is all data of the world is dirty and our system assumes that the data is going to be dirty and install into their dirty data. But we don’t expect you to sort of massage data in a certain way, or create a particular very specific kind of schema and set it up nicely. If you want to sort of apply a cleaning filter on top of your data so you might get very dirty data and then we allow you to put a simple filter which is a matter of hours and then you can define that and then expose that to your end users. And we will basically go to filter so all the dirtiness is filtered out and we are presenting clean results to end users.

So the technology that we have built, the search engine, that leverages all the investments you have made in your data sources already to cut down that time and we don’t have to go through the long process of defining any kind of natural language models or any machine learning models and then test them,verify them, and over time we get better – that’s not the model. On day one we are good, over time it become great.

Martin: Ajeet, how did you come up with the revenue model and what is actually driving the pricing? So is it more like storage, or computational power, or the number of users, or the just companies size? What is driving the revenue?

Ajeet: Yes, it’s a good question because if you look at the pricing model that exists in the market today it’s a per user model. So if you’re a customer and you’re signing up I would ask you: “How many users do you think you will have?” And you say 100 and I’ll try to sell you license for 200. And he’ll say: “Give me 150. Let’s started with 150 and will see over time where we end.”

And then IT Department says: “We have 150 licenses for this.” First of all, since the technology is complex lot of those licenses go waste. And if more people want to access to information, let’s say you go to new department which has 1000 people, then you need to go back to the vendor and again ask them to spend more money with them.

The whole idea behind ThoughtSpot is adoption of data. Because today only 22% of people can access data and we want to increase that number significantly. And for each of those people, we want them to be able to access the data very-very frequently, so we are talking about 10x to 100x increasing adoption of data.

So we don’t want to penalize our customers on number of users, we do not charge based on the number of users. Even our smallest product skill that will sell to you, you can put unlimited number of users on that. How we price our product is we set up lines and it’s building block models, Lego block model. So you can drag and stack them as much as you need to scale and you can start small and grow over time, but we do not have any limit on number of users for these appliances.

Martin: Is this then query based or data storage based?

Ajeet: It’s based on appliance which can find amount of data.

ADVICE TO ENTREPRENEURS FROM AJEET SINGH

Martin: Let’s talk about your advice for first time entrepreneurs. This is your second company. What have you learned along the way where you said this was a major learning for me going forward and I will never forget this lesson?

Ajeet: Yes, there are lots of them. If you talk to anybody who has being at the startup at any capacity it’s the best place to learn. So my number one advice is to find amazing startups and spend some time there either in a full capacity or an employee. Because if you’re someone who’s driven by passion for creation which is most of people are at least here in the Valley and many other places in the world where large technical talent. So when working at the startup aou see the opportunity to have a much more open playing field where you can have your ideas come to life very quickly.

If you are starting as an entrepreneur I like not to be very prescriptive about these things because every situation is very different. But for me I think what is important is understanding the risk model you’re opting into, either directly or indirectly. I like to look at risks in two dimensions: market risk and execution risk. Is there going to be a market for my product, the idea that I have? And that is market risk. And if I am actually successful then i is going to be valuable that’s how I look at market risk. And I also look at the history of companies that have become successful in this market. And if I see a market that has 10-20 billion plus dollar companies built in the last ten-fifteen years that tells me that this market is large and it supports building large independent companies. On the execution side it is – Can I actually build the project that I’m talking about and sell it? That is the execution risk.

So I personally like to go after opportunities that are very low in market risk and very high in execution risk because I don’t want to spend several years of my time and some of other people time and find that there is no market for my product. But we want the execution risk to be high because you want to set a high bar for anybody else to be able to copy what you’re doing.

So that is one model and I’m not saying that should be the only model. There is a way of building companies that falls in this bucket. There is another model that is high market risk and low execution risk at least to begin with where I don’t know if there’s a market. If you think about Uber or something else, it’s questionable whether market risk is low or high. But lots of consumer startups would come in that bucket.

But what you have to focus on is really mitigating market risk before you do much execution. So understand the risk profile you opting into and focus on that part first. If your execution risk is high then mitigate execution risk by thinking through the product you’re going to build and thinking how you will set it, thinking through the architecture of the product before you start calling. If your market risk is very high get the hell out of your office and go out and understand what the market is like, prototype and understand all the lean startup stuff.

Martin: And assume you would have checked the market risk which is low and now the question is if you’re checking on the execution risk how do you identify whether you potential solution will be like 10x or 100x better than existing solutions?

Ajeet: I think some of that comes from what people call it vision. It’s like you understand the market and you understand the problem. Now I was saying early in the interview, the way I like to think of the idea is – what is the most awesome way of solving the problem? And the most ideal way if I had infinite money and infinite amount of people, everything? What is the most ideal way of solving this problem? And then go from there, then you start to make it viable and fix the execution risk. So I would say in the beginning it has to be driven by your vision and passion and gut. There is no sort of science to it, big part of building a company is making gut calls.

Martin: Ajeet, thank you so much for sharing your knowledge.

Ajeet: Thank you, Martin. Thank you so much for having me.

Martin: Great! And if you have a company and you don’t know whether your business people are really checking all the data and getting the most out of, check out ThoughtSpot. Thanks.

Comments are closed.