The following is a transcript from a podcast interview with Brad Peters, Birst Chief Product Officer by Entrepreneurial Insights’ Martin Luenendonk:

Martin: Hi, data is so much around us but the major point is we need to find insights in it. Today I am here with Brad. Hi, Brad, who are you and what do you do?

podcast1Brad: So I am Brad Peters. I am one of the co-founders of a company called Birst – BIRST. We are a cloud based business intelligence and analytics company focused on helping organizations take data from their operations in businesses and help them make sense of it so they can run their businesses more efficiently and effectively.

Martin: Cool. When did you start this company and how did you come up with this kind of business idea?

Brad: I started in 2005. I have actually been in the analytics space for some time. Interesting enough, prior to starting the company which was really based on carrying some of the things I had seen in a prior the prior life forward into what we saw was a more modern era. Prior the company I was actually in another company called Siebel systems which was then a large customer relationship management software company. The company that sold solutions for sales, service and marketing organizations, intended to have a lot of people that use their software, arguably the predecessor of

We discovered something in the late nineties that all of this customer data was going into systems and into our system and we were doing a good job helping sales reps put stuff in the system but we weren’t making good use of that data for the purposes of managing the business or understanding our customers or doing anything like that. So we decided to embark upon a journey of seeing how we can make this data more useful. And we did that by partnering with some existing business intelligence providers. At the time the business object was for example the partner that we chose to use and put on top of the Siebel to see effect that would work for us.

We tried it but what was interesting is that was a product that was built for relatively limited use years and years prior and our customers really had challenges using it. It was challenge product line. And one of the challenges was that unlike how most people had used these other types of analytics products before in the past which were usually a few people at a time at a department who were super technical, we were selling to sales people who probably didn’t like tech. Technology wasn’t a big part of their skill set and we wanted thousands of people to be able to use information and data and that really wasn’t how people thought about analytics and data before. And so we were challenged there and we had to come up with a solution.

So we took a second try at it and we ended up buying and building some technology that was really about how do we take analytics and spread it out to a lot more people in an organization. We created, if you are a technologist at all, when web servers first came around there was this technology called application servers that were designed to build scalable applications delivered to a lot of people around the web. We kind of build the first one of those from analytics and we saw that really succeed very, very well. In fact, the analytics product line at Siebel became the largest product line in the company over the next several years.

It really spoke to a couple of things. On the positive side it spoke to this incredible demand by regular people that has been growing as far back as I remember to have access to facts and information to make decisions.

You know probably 30 or 40 years ago it was generally accepted that you made decisions based on the rules of thumb, habits or things like that, but I think this is really accelerating in the last several years. Even ten years ago we were seeing that people were much more comfortable making decisions and there is the much greater desire to make decisions based on facts. And so the demand for our products was increasing.

Maybe less positive thing or slightly negative thing was the other products in the company were shrinking so we kind of crossed in the middle why were the other products shrinking. They were older technology; they were built on what we would call it client server technology. They were not web or cloud based. And we were seeing those products being basically disrupted in the market place by other cloud providers, namely guys like salesforce and folks like that. The big advantage or the many big advantages of the cloud is the fundamentally new way of building and writing and delivering software than it had been done in the past. As a consumer it is just a lot easier to consume the cloud way less painful, way more friction free and so people were moving that way on the CRM side.

Martin: Did you start Birst as a cloud service provider already or did you just come one or two years later?

Brad: No, so this is the thing we said “Okay, if this CRM stuff is being disrupted by cloud, by guys like salesforce and right now and omniture and you know, go down the list. And because it is hard to install, difficult to maintain and all this nonscalable, all this sort of stuff well shoot, analytics is even worse, because there are even more pieces to put together when you play with analytics. Maybe the cloud has a role to play for analytics”.

So we started Birst in 2005 with the vision of bringing analytics to a modern cloud based architecture. I think in hindsight we were probably a few years ahead of the market when we decided to go do that. But yes from the very beginning we said: ”Look, there are major architectural shifts that go on in software probably every 20 to 30 years. We are seeing one right now when we went from main frames to minis, from minis to PC, to client server, to now web. We have seen these massive shifts. So whenever there is a massive shift there is an opportunity to rebuild and rethink, reimagine if you will the prior generation of stuff that came before.” And we set out to do that in the world of analytics.

Martin: Brad, imagine I am a company and I have got lots of different data sources like Google Analytics, I have my own web logs and maybe some API data and so on and so forth. How does it work? How do I get this data into your kind of Birst cloud platform? How do I get some analytics out of there? And how do you make sure that the quality of the data is ensured?

Brad: Great question. The interesting thing is that this is what the hard stuff is. That is what most people who don’t come from an analytics background easily mistake is that they look at pretty pixels on a screen and they say: “Oh, it is a pretty chart. That’s where the value is.” Reality is I think that the charting and the visuals while pretty are fairly simple. That is not where the hard stuff is. That is not where the value is. The value is in the data. It is in coming up with answers. We like to say at Birst that pretty wrong answer is still a wrong answer. It is all about how do you create an infrastructure so that you can get the correct answer or you can get the answers that you need to the questions that you have when you need them so you can make decisions based on facts. It turns out that is not easy to do. That was another thing that we kind of even as analytics veterans we underestimated that because it is extremely hard.

So the challenge is even more broad than just say Google analytics and some web log data and things like that. Most companies that we deal with have that. They also have Salesforce they have a bunch of stuff that is inside of their firewall on premise or they may have a data warehouse already. They have a bunch of stuff sitting in a bunch of different places that each give you a silo or piece of information about how their business is performing but the question they want to ask span those silos. They want to ask questions like when I did that web advertising campaign how did that turn into leads and did those leads close into deals and how much did it cost me to generate a customer? Those are pretty expensive questions that you can’t answer by taking one of those pieces by itself. You have got to look across all of them.

So we had to spend a bunch of years building technology that can handle data in two ways. The first way is we can take data and connect to something like Google Analytics or Salesforce or SAP and we can extract data and we can make it what we would call analytically ready because the applications in its raw form not really good for answering questions. It is built in ways, there are whole ways that engineers structure data for the purposes of application that make it hard to use for analytics. We turn it into an analytically useful form.

But also, there is other data that is sitting out there that is already been worked with and is tuned into something that is useful in which case we don’t load that into Birst. We just connect to it. We map on top of whatever it is and then when we need it we just query it in place and so we create this layer, we call it our user data tier, and that basically allows us to present to the end user this integrated picture of all this data in their company. Even if some of it is in Birst and some of it is not, we created this unified view then we can then allow people to ask questions off, create visualizations and dashboards and reports in a whole variety of ways of looking at that data so they can ask and answer the kind of questions that they want.

Martin: What happens, Brad, if I am having like you said different data sources but in the history I wasn’t aware of that and I was only looking at the silo type of analytics which we both agree is not where the value lies. I am pumping the data into the Birst platform but apparently how do you want to join this data if I don’t use the same kind of user identifiers or different time stamp technologies or something like this?

Brad: It is a great question and I think this is where a lot of people get hung up with analytics. So in these different silos I would say in our empirical experience more often than not there are relationships between data that can be exported directly and this notion of a customer name being different in one place and being different in another place. While that is true that particular issue is a smaller issue than we typically see in larger systems and there is a ton of value that can be gotten out without solving those types of fuzzier issues. Out of systems just straight as they are with a little bit of extra work we can tie those systems together so they generate common identifiers and do the kids of linking that you expect. It is not magic and that is something that folks need to keep in mind. But also it is not instrumentable either. It takes a bit of work and there is a well-defined best practice and by doing it intelligently you can minimize the amount of work involved. There is still some work that needs to get done every time you want to bring in a new silo into your overall mix in terms of how that silo relates but through intelligent use of automation and other types of tools we can keep that as a manageable piece of work.

And then the benefits of once you have done that, once you have created a mechanism for cross keying various systems or relating these different elements. Keep in mind relations can be as low level as I have a transactional key that synchronize across different systems. They can also be as simple as time. What if I just know that have spent so much in advertising revenue in a month and I have got so much in leads. That’s valuable in on itself and certainly not an excuse. You can conform data on multiple levels and you don’t have to solve the intergalactic data integration issue to get a ton of value out of it. I think the goal of analytics is to do everything incrementally and do it iteratively and start by taking the lowest hanging fruit and continually to take more and more chunks of value off the table as you continue to add more richness to your data set. But not having a perfectly integrated data set is not an excuse for not starting.

Martin: And how do I ingest all my data sources into your system. Do you have APIs for all systems or do I need to build some kind of data pipelines myself?

Brad: We do. So that is one of the other challenges we had to solve when we moved to cloud. We couldn’t assume and in some cases it would have been ok to assume but we didn’t feel we were in position to assume that all of the data pipeline and data integration and data transformation logic would be done before the people gave the data to us because it would be wonderful if everybody just piped into Birst a super clean single table that added everything exactly as we wanted and all we had to do was chart it. I don’t think in the history of Birst that has ever happened.

So we actually have a data pipeline as a part of our process and what we wanted to do is not just add the data pipeline but have that data pipeline be built into the visualization and analytics pipeline as well so that when you define something and bring it in by connecting it to the other piece you minimize the amount of duplication, integration work you can automate a lot more of those hand offs by having it be together. So we have APIs that can get data all over the place whether it can be cloud based services thing, like RESTful API and those sorts of things or on premise databases or file systems, or Amazon, or all the various ways that SAP has you to connect SAP. All of that stuff we have connectors for. We have built hundreds of connectors for hundreds of types of ways to bring data in and connect to data.

Martin: And how did you start. I assume your prioritized those connectors over each other because in the beginning you did not have infinite resources. How did you prioritize in the beginning?

Brad: Really good question and this is the standard product management question where you had to be ruthless in your prioritization about how stuff works and what you are going to do.

When we got started Salesforce was not that big. We were tiny, but they were not the dominant factor that they are today. So the biggest single source of data that we actually saw was databases. The good news is some of that has already been standardizes so we started out with that and we started doing the standard exercise which is let’s look at the market and let’s figure out some combination of presence where is most data in companies that we see. And we line that up with what kinds of customers do we thing we were initially going to get and where do they have most of their data. And that helped us prioritize and come up with a few some early sources and then over time we built an infrastructure that allowed us to not have to build these in one of fashion that allowed us to build a framework for adding new sources that then we can scale up and turn into a connector machine if you will.


Martin: Brad, when we are thinking about business model what are typically your customer segments and can you give us some kind of demographics or statistics on them?

Brad: Yes, sure. I would say it has changed over time. So where we are today is Birst serves more often than not larger companies or more sophisticated analytical or data scenarios. If you have a single table and you just want to put a chart on it there is a hundred different things that you can do to go do that. But if you have hundreds of tables or tens of tables coming maybe from many different applications we call that schematically complex data. And when you have schematically complex data that is where Birst shines. We are much better at handing that than other tools. And you typically find that data in larger companies with more sophisticated environment or in organizations where you are taking a large percentage of what they are doing and making that analytically useful so say that for example embedded use cases. Somebody is building an application, they want to add analytics to it and they will embed us in the process.

We sell right now to other software companies that embed analytics in their application and we help them get their arms around that data and then we also sell to companies for use internally to analyze data from the applications that they are using. So one is for companies to resell to their customers the other is for companies to use internally. And for the internal use case like I said it is moving to larger customers but I would say five or six years ago I think the cloud was probably a little new for a larger customers and larger customers were still playing around with a lot of the legacy vendors – the big mega vendor guys and were just learning about the cloud. So we were probably focused more on what we would call the mid-market – mid-sized companies who probably didn’t have any analytics software and we became a one stop shop for these folks, because we had everything. And over time as our product has gotten even richer we have seen those customers get larger and larger.

Martin: Cool. Brad, how did you find and acquire the customers in the beginning? So imagine, after you have built the first integration of your product, just trying to go to the market. How did you find and acquire those customers?

Brad: We’ve got to several go to market models. The initial go to market model was actually different. Initial go to market model was predicated on the idea that it was going to be very hard to do a mass market approach for small company. So let’s not do a mass market approach initially because that requires a lot of capital and it requires a lot of brand presence and those sorts of things. It was something that’s a lot more focused.

Initially, we didn’t have a lot on the BI or analytics piece and so what we decided to do was actually to build the applications rather than to build a full line analytics product. We said: “Let’s go after a single use case. Let’s go after a single application for a single industry in a single vertical” and in our case financial services and wealth management and we said: “Let’s build custom built solutions around that.” And when we do that we go from company to company within the vertical and as we do that we will build out our technology platform and make it more horizontal over time and then we go to another vertical or two. And then eventually we will go horizontal once our platform is built out to be able to be marketed directly as a platform.

So we actually ended up targeting the use case and application first and didn’t have much product actually built initially. It was more around work: “Here is an area of value for you Mr. Customer. If we are able to do this clearly this would be of value to you, do you agree? Yes. Ok, so let’s go jointly and get building this together.” So they were a relatively small number of people we found a couple of early customers that were partners with us to build those early use cases and we used that to push our platform forward.

And then as we push the platform forward we made It more general, we got better and richer. And then we get a big push when the financial markets imploded in 2008 and a lot of our target customer base kind of went away literally. And I said: “Okay, if our core market is gone and the next two or three markets that we are going to go sell to are gone maybe it is time to go horizontal now. ”

So in 2009 it was when we shifted to horizontal. But by then we had enough critical mass that we had from salespeople we had some marketing people we can go and do the kids of things you would expect selling to more horizontal set of capabilities would need.

Martin: And how is the revenue model working? Especially how do you price and what do you price.

Brad: Well we are consistent with most cloud companies. We are typically per user per year. We have customers that sign up. There is generally a modest startup fee just to get folks kind of hooked up but as a standard cloud model which is the more you use it the more value you get, the more we participate in that value.

Martin: Because I would have assumed a hybrid model for example one part being on a per user model or per user, per month whatsoever and on the other hand based on storage.

Brad: The thing about storage is we do have to charge for it because if there is no charge for its customers can just go crazy and put infinite amounts of data in there. At the same time storage has become so commoditized and storage is one of those things where this is the marketing challenge we run into which is when we store data we don’t just store data on a disk. We have high performance database that is tuned for analytics, that is an analytical system with high performance hardware around that. It is designed so that when you query that data you get fast results and you get the kinds of things that you want. That is not like Amazon S3 where I am sticking something on a cheap disk somewhere and I may call it once in a blue moon. This is very, very different environment.

So the challenge we have had is it is more expensive than more disks. But when you charge for storage the average instinct of most customers is: “Shoot, I can go down to fries or go on whatever website I want and I can go buy a disk drive for 200 bucks that contains terabytes. Storage should be free.” And they don’t think about it is not just the disks or the storage. It is all of the other stuff that goes with it. By the way we have to back it up and we have to have disaster recovery and reliability and all the things you have got to do. But it is still just instinctively for customers tough to think storage is expensive.

So what we have done is we have made that as small as we can and so typically is part of our customer model unless the storage gets really big. The storage is actually relatively small piece of the overall picture. It is more based on the value in users.

Martin: Ok, cool. Who do you perceive your competitors to be and why do you think you are better and in what dimensions are you outcompeting them?

Brad: We kind of have two classes of competitors I’d say in the market place. Ultimately, what we are trying to do – we are trying to create a disruptive way of thinking about analytics in the market place to give people better economics and better responsiveness and better ability to make decisions with data. And so really we have to compete against the alternatives the people are looking at in the market place. And there are kind of two alternatives the people see.

The first is the legacy toolset that people have. They are typically the mega vendors that you can think about – the big blue and red and other logos you can imagine. And those folks have toolsets that have grown up over the last 25 or 30 years. The core code for these systems was probably written in C++ 25 years ago. So it is old stuff, really old stuff. Analytics is a complicated beast and so there is a lot of different pieces to building an analytical system, we were talking about data pipelines and those sort of things. When you go to these companies in order for them to give you a solution it is not one product. It is probably a collection of at least 5 to 10 different products that you have to have and you as a customer have to like put it all together and make it work. We like to say the only thing integrated about these like legacy guys is the price list. You have big price list that has lots of stuff on it. But then once you buy it you have got to go put it all together and that is one of the big areas where people don’t like analytics because you call up the IT people and say: “Hey, I want to get analytics on this data” and they say: “Okay, we will see you next year because we have got to go do this major project that is probably similar to building an aircraft.”

So that lack of integration makes them there is a lot of stuff you can do with them but they are really low level and they are leveraging really old components. They don’t get any benefit or any lift from any modern software architectures and techniques that are being done today like RUIs is based on the Google framework and we have got stuff from Facebook and Netflix and stuff like that built into our product. They get no lift from that, they are completely fragmented, they are expensive but they have a large vendor relationship because they are big folks and a lot of people don’t look really deep when they buy the stuff. So that is one opportunity for us to radically change the economics of analytics by making stuff way easier to consume within those folks and that is one set of competitors in the market place.

The other set is what we would call desktop software. Over the last 6 or 7 years there have been an array of new folks that have shown up in this market place that have said: “Wow, these suits of tools that the large enterprises have collected over the years by acquiring all those companies are really complex and so what we are going to do is we are going to focus on a much simpler problem.” If you only have a single table or maybe a couple of tables let’s at least make that really easy to use one product and let’s do that on desktop. So anybody anywhere who is an analyst can plug a data on the desktop and make a pretty dashboard or pretty chart.

That works really well and they have done really, really well with these limited use cases and they are doing that certainly next step up from where maybe Microsoft Excel left off. Excel is probably the most prevalent analytical product in the world and not probably, it Is way more prevalent than anything else but it has it’s great deficiency in that the format of the data you are looking at and the data itself are one and the same. They are both in the same place. So if I create a report today with an Excel and I want to do it again next week I kind of have to rebuild it from scratch. There is no scalability or leverage in Excel. And Excel is really hard to manage and so it is the first thing but it is really messy.

These desktop tools kind of go a little bit beyond Excel and they let you take some tools and create more repeatable process on the desktop level but they gave up on enterprise and sophisticated data sets for bigger problems and bigger analytical needs. And so where we fit is really when you have something that is more than just a few individual or small problems, you have something that an organization needs to look at when you want to deal with organizational level analytics and you don’t want to endure the pain and suffering associated with these legacy tool sets were built in prehistoric year. Now you have an opportunity to leverage cloud architectures and those sorts of things and handle those problems. And that is where we fit, it is really agility of some of these desktop tools but on an enterprise or organizational scale.


Martin: Good. Brad, let’s talk about your learnings during your entrepreneurial journey. What have been your major learnings and maybe your biggest mistakes?

Brad: Well, I think learning occurs on a couple levels. I went to business school I had a technical background coming in so I grew up in an entrepreneurial family. My dad started a company in the late 70s around software for the early microcomputers that were there. So I had some exposure to what starting a company was like and starting a company in San Diego in the 1980s was not an easy thing to do. There wasn’t a lot of venture capital or anything like that around so it was really hard. So I had appreciation going for this stuff. You could jump on a gusher and there are a lot of people that mistake luck for brilliance. They tend to assume they have much higher IQ than they actually do because they got lucky. That is possible you can certainly get lucky. You can certainly find an opportunity that with passable execution will get you to something that is successful and those happen and they are stories in the Valley and they become these legends that people get excited about.

But I think the much more common case is this stuff is pretty hard to do because nobody really wants you to succeed. I think they like the idea of the new guy but nobody wants to buy from the new guy. So it is much harder to get going. And so I kind of knew that going in and my partner was in a similar mode. His father started some companies as well. He actually founded Polycon and Picturetel and a number of companies that are pretty famous. I would like to say one thing.

When we started he gave us one word of advice. He said pretty apt: “Whatever you do, avoid death.” And that sounds really trivial but in reality a lot of people make a lot of moves with this idea that being bold is great but more companies die then don’t and if you do something that allows yourself to be killed you don’t give yourself the opportunity to be in the right place in the right time when the market moves your way. Most companies that are successful are there when the market is there. You can’t really move the market so you have to be around until the market is there.

For us what I think we focus in learning was learn how to quickly read the market as best we can and try to adapt to what we are learning as quickly as possible. Our market has gone through probably three or four major changes since we started the company and we had to change three or four times pretty significantly. And it is always a balance, you want to jump very quickly on the new thing but there is risk associated with that. So it is how do you walk the line of putting a foot forward on the next rock that you are going to put your weight on but don’t take your weight fully off the rock that you are currently standing on. But it is sinking, it is going away so you have got to make sure the next rock is stable before you put all your weight there and keep your balanced. That really has been our focus – making sure that we make fast but judicious steps forward.

Martin: You nicely put the frame when you said that the father of your co-founder gave you type of advice. What advice would you give your children if they wanted to start a company?

Brad: Fail fast. I love the avoid death because it has been true in every case that I am aware of but as a second piece of advice I think – fail fast. It is very easy to hang on to something that is marginally successful but won’t get you there and if it is not working admit failure quickly and move on.

Martin: And how do you identify whether it is failure or whether you just need to stick around a little bit longer and waiting until the market keeps moving?

Brad: That is the big billion-dollar question. I think that is where this whole idea of being adaptable is trying to figure that out as quickly as possible. And that means being ruthless in being honest with yourself are you failing or not and sometimes you have to stick in there. You have to say: “Look, I don’t know if this is succeeding or failing and so here is what I am going to do, but these are the signs I am going to look for and when it comes to this we are going to fail”. But don’t fail on an idea before you get the next thing lined up because you don’t want to shoot your foot off.

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

Brad: Thank you I really appreciate it. Thanks, Martin.

Martin: And if you are looking for a great cloud BI solution and you have a lot of data scattered along your company, check out Birst. Thanks.

Listen to the original podcast.