How the right tools can create data-driven companies, even at Facebook

2 Comments

Analytics startup Interana launched last week promising a product for analyzing event data that marries a super-scalable data store with an easy visualization layer. Gigaom covered the launch, touching on the company’s Facebook roots and its theory on why event data matters, but co-founders Ann and Bobby Johnson also came on the Structure Show podcast to delve a little deeper into these topics — and to explain why starting a company with your spouse can be not only rewarding, but also a wise business decision.

Here are some highlights from the interview, but do yourself a favor and listen to the whole thing. You’ll get co-host Barb Darrow’s and my thoughts on the big HP breakup, and then the whole story from the Johnsons on why they launched [company]Interana[/company], how companies are using it, and all that’s possible when you can track behaviors of anything from servers to customers over years.

Download This Episode

Subscribe in iTunes

The Structure Show RSS Feed

Taking Facebook’s experiences to the mainstream

Bobby Johnson, Interana’s co-founder and CTO, spent nearly six years at [company]Facebook[/company] and led its engineering team. Interana’s third co-founder, Lior Abraham, was also a Facebook engineer and was responsible for a popular analytics tool called Scuba. Here’s how their experiences at the social network giant informed the product design at Interana.

“We saw a lot of data. We built a lot of things to handle large interconnected data and try to make things really fast,” Bobby explained. “From that experience we just saw a lot of things that we could do better in the analytics space, so we’re taking a lot of those things that we learned from scaling Facebook and applying them to this problem.”

One of the things he saw was how fast the right tools could spread from their initial users and uses into entirely new areas:

One of the first things I wrote at Facebook was a program called Scribe, which just collects logs. And when I first wrote it I thought we had four use cases we were targeting and we could think of another three or four we thought might be interesting. So we stood it up and we made it really easy to stick in other datasets, and within six months we had more than 100 datasets. It turned out that people want to know the numbers, we never had a problem convincing people to care about the numbers. The problem was just it’s often sufficiently hard to get to them that people just kind of give up and they make a guess instead.

Ann Johnson answered on behalf of Abraham and his experience building Scuab:

He put in a lightweight backend that let you save two weeks worth of data, and people loved it. He thought it would be used by the performance team, but within months it was being used by all of engineering. And then it moved not only to engineering, but to customer service and HR and marketing, and they were all using the same simple interface.

The Interana founding team. Source: Interana

The Interana founding team. Source: Interana

Rich data plus timestamps equals behavioral insights

“[N]ow that storage is very cheap, people are saving richer data, and what that rich data looks like is just a series of events over time,” Ann explained. “And the reason people are saving it is all of a sudden that opens up looking at behavior. So you can look at the behavior of your users, you can look at the behavior of your products. And the cool thing about behavior is that’s where a lot of your really interesting business metrics are.”

Asked about the difference between Interana and something like Splunk, she added, “The thing about machine data … that data kind of gets boring after a couple weeks, you never really go back and look at it. When you’re looking at user behavior data, you look at it over 10 years, you look at for as long as you can.”

The Facebook effect is already happening

Interana’s beta users include a list of big-name customers largely in the web space (Sony, Jive, Asana, Tinder and Orange Silicon Valley, among them) and they’re already doing exactly what Interana hoped by deploying Interana for specific use cases and then quickly finding new ways to use it.

“We call them ‘off-label’ uses,” Ann joked. “… The performance guys are always hungry for more data. The performance guys always run off in the corner and start using it as well. That’s happened in at least half of our beta customers.”

She added, “We’re trying very hard with our pricing to enable this kind of viral spread throughout the company, because that’s what we saw at Facebook and that’s what we’re seeing at our beta customers.”

A screenshot of the funnel analysis in Interana.

A screenshot of the funnel analysis in Interana.

I asked whether that might not be easier with a cloud service rather than enterprise software, and Ann said Interana was initially planning to launch a cloud service (and still easily could) but was brought down to earth in part by customers’ concerns over having to move too much data.

“We were so sure we were going to do this in the cloud at first. We were like, ‘Salesforce did it in the cloud, everybody’s doing it in the cloud, we’ll do it in the cloud!'” she said. “But as soon as we started out doing some of our customer research, everyone was like, ‘No. We will not talk to you if you’re in the cloud.'”

The family that wire-wraps motherboards together …

Technology aside, the most notable thing about Interana might be that Ann and Bobby are married — a rare thing among tech startups. To ask them, though, it’s not weird at all, especially if it’s indeed true that investors look for good teams as much as good ideas.

“It’s wonderful,” Ann said of the startup experience so far. “We were at Caltech together, we used to wire-wrap boards together, we’d build crazy stuff for parties. We love building stuff together, we always knew that. It was always our dream to work together, but we never seemed to get it to work when we were working at other companies.”

That kind of relationship is hard to beat, Bobby said: “One of the things about starting a company is you have to have people you really trust and really know you can work with, and we’ve been doing projects together for 15 years. I can’t imagine starting a company with somebody I didn’t know that well.”

2 Comments

Dominique Levin

Great article. And they seem like a great team.

Relevant behavior doesn’t always need to have a timestamp. There are many other descriptive factors that we use to predict shopper behavior that doesn’t need a timestamp. For example, SIZE of first order, NUMBER of product types in cart, ACQUISITION source, etc. (here’s a list of ten more: http://bit.ly/1BJn3Px), don’t need timestamps to predict behaviors. These non-timestamped metrics don’t seem like they could be easily integrated with their platform based on the screenshot of their app.

It would be interesting to see how they incorporate these metrics.

Comments are closed.