More B2D: TappingStone Wants To Tap The Need For Data Scientists With Its User Behavior Predictions-As-A-Service

Next Story

MongoHQ Raises $6 Million For Database As Service Engine

Startups catering for developers, so-called B2D, are undoubtedly on the rise. I’m starting to lose count of the number of times I’ve been pitched with something along the lines of “like Twilio but for X”. And it makes sense. With money sloshing around the Valley and elsewhere, and conversely, a shortage of available engineering talent, offering startups the ability to outsource some of their technology would seem a business opportunity too good to miss (a point well made by Semil Shah in a recent TC guest post).

Hoping to seize that opportunity is London-based TappingStone, which has launched its user behavior prediction-as-a-service in private Beta. It’s promising to enable app developers to build smarter apps by utilizing the company’s API and backend algorithms. Based on the data that TappingStone then returns, developers can personalise their app for each end user, and in doing so, not only improve the user experience, but also up sell in a more targeted way or do other interesting things like lesson churn by predicting which users are less likely to stick around before they leave for good.

These are just two, fairly primitive, examples. But, essentially, developers plug their user data into TappingStone and then query its API to get user behaviour predictions back. What actions they take based on those predictions is up to them, but with TappingStone doing much of the heavy lifting, it opens up all sorts of possibilities.

“We believe the demand for a service like TappingStone will be huge for a number of reasons”, says co-founder and CEO Simon Chan. “First, big guys like Facebook, Google, LinkedIn, Zynga, Flipboard etc. are spending a fortune in data engineering. Apps have to be smart nowadays. Consumers just expect so. Secondly, with the growth of smartphones and tablets, people consume more apps, and more people want to develop apps for this growing market. Thirdly, despite the need, there is a serious shortage of data scientists and engineers. Startups can hardly recruit the people they need, while many companies feel the pain of retaining talent.”

It’s a decent pitch, though it would be remiss not to acknowledge that there are competitors. PriorKnowledge springs to mind, while Chan brought my attention to Precog. And I daresay there are others, along with open source recommendations engines and algorithms from which developers can build on.

Where Chan says that TappingStone differs, however, is that it doesn’t take a one-size-fits-all approach. Instead, he claims that it’s able to construct a unique algorithm adaptively based on each specific dataset and user behaviors.

“It is technically challenging to provide such self-served behavior prediction service for all kinds of apps. The needs of behavior prediction for, for instance, a news app and a restaurant discovery app are totally different: one is very time-sensitive while the other is more dependent upon location”, explains Chan, whose background includes building various social startups in Silicon Valley, Hong Kong and China, all of which required developing ‘smart’ features, part of the inspiration for TappingStone.

I asked Chan to provide a few tangible examples of the types of apps that could utilize TappingStone to add smart features.

One example he gave is a mobile game based on a freemium-model games which could use TappingStone to predict which players are more likely to become paid customers. They could also increase revenue by offering personalized in-app purchase discount offers to those customers who would be most likely to convert as a consequence.

Or a travel app or hotel booking app could do something similar to Amazon — “People who have been to Paris also like Munich” — along with recommending things to do and see, the next trip a user might like to take or even predict when a user will want to go another trip.

And so on.

It terms of how TappingStone plans to make money, it will follow a similar model to Twilio and charge developers per-usage, priced per data size and per prediction calls. I’m assured it will be cheap or possibly free for smaller startups but will scale competitively for something like a Pinterest.

“We haven’t decided the exact price range yet. But it’ll be dirt cheap, especially compared to in-house data scientists”, says Chan.

TappingStone is currently self-funded, while its CTO can’t be named as he is currently seeing out his notice period. You gotta love startup life.