Lessons learned: How to get your big data startup to a Series B

Lost in the news of Appcelerator’s acquisition of Nodeable on Wednesday is a lesson in how to build (or not to build) a big data startup. Nodeable Founder and CEO Dave Rosenberg explained that although he and his team are happy to be part of Appcelerator, an acquisition this early wasn’t part of the plan when the company launched in 2011.

Every venture capital firm has made some preliminary investments in big data companies, he said, but those companies will likely need some recognizable adoption or income in order to justify further funding on desirable terms. Even for a team of veteran entrepreneurs and technologists like Nodeable, which had built a platform that served as a front-end complement to Hadoop by analyzing streaming data, it can be challenging to get over that hump. We’ve written about this in our coverage of how challenging it can be for companies to get their Series B round of funding without having transcended to rockstar status.

Dave Rosenberg (left) speaking at Structure 2012.
(c)2012 Pinar Ozger [email protected]

Despite not making it past the first round, Rosenberg has learned a lot. Here’s what he thinks he’s learned about how the big data market is shaping up and pitfalls others startups should try to avoid.

1. If you’re doing big data, do it open source and on-premise.

Nodeable thought it was getting ahead of the curve by launching its stream-processing platform as a cloud service, but it probably was too far ahead. With rare exception, Rosenberg said, “Organic growth in big data right now is clearly with open source projects that are behind the firewall.”

Everyone agrees in theory that delivering big data tools as cloud services is the right way to go (myself included, for what it’s worth), but most companies aren’t there yet. Even with Amazon’s popular Elastic MapReduce service, he added, people are running a lot of jobs but not necessarily only exposing small amounts of their overall data sets to the cloud and then are pulling the results right back behind the firewall.

“From a user perspective,” Rosenberg said, “they want to download things.” That’s why, he added, “Even though Amazon [Web Services] is booming, it’s still websites.”

So, rather than hire an enterprise sales team to start pushing its product,¬†Nodeable’s next move if it didn’t sell was to open source its technology so the large companies that expressed interest could begin experimenting with it on their own servers. That wouldn’t have guaranteed success, but it would have aligned with companies’ realities about where their data is stored and who has access to it.

2. Hadoop is not a license to print money.

Although some companies are making a lot of money selling Hadoop product, simply incorporating the technology into a product does not guarantee success. “Cracking into the Hadoop space is very difficult,” Rosenberg said.

Nodeable, for example, had close relationships with key members of the Hadoop community — including Cloudera CEO Mike Olson, who sat on the company’s board — anf it still wasn’t able to tap into that market like it had planned. Part of that ties into the previous point about being an on-premise solution, but another part has to do with technology choices that align with where others are pushing the platform. Whereas Nodeable relied heavily on Storm for real-time processing, the Hadoop platform vendors such as Cloudera and Hortonworks are putting their real-time energy behind HBase at the database layer.

3. Life is easier as an app.

In some ways, Rosenberg said, companies that sell applications might have an easier time succeeding time than those that sell infrastructure. The former come with their own use cases, while the latter are always searching for new use cases that let them find their way into new markets. In cloud computing, especially — where Rosenberg says people often expect free software — it’s probably easier to get them to pay for applications than for big data infrastructure. (Nodeable, I would say, fell somewhere in the middle.)

And when acquisition time comes, cloud companies such as Salesforce.com (s crm) and Workday (s wkdy) that have to make big data plays will probably be looking for applications rather than infrastructure tools, Rosenberg said. (Workday, for what it’s worth, just partnered with Hadoop BI startup Datameer on this front.) If you’re looking for the best-possible situation and don’t want to end up part of a company that will take your talent and bury your technology because it thinks it can build something better, having an army of loyal users is good leverage.

4. If you must sell, choose wisely.

Appcelerator wasn’t its only suitor when Nodeable decided to sell, Rosenberg said, but it was the best one. Rather than being an “acqui-hire” situation, where the hard work the team put into Nodeable’s StreamReduce technology would have been lost, the technology exists and will help build a real-time analytics engine into Appcelerator’s mobile-app development service. One reason he’s so confident about the decision is that Nodeable already knows the Appcelerator team and its plans — the two companies had been working together for months to improve Appcelerator’s infrastructure.

Feature image courtesy of Shutterstock user Rashevskyi Viacheslav.