Tuesday, April 30, 2013

Business model for MLaaS - Discussion at Data 2.0 Panel

I had the opportunity to sit on a panel aptly named “Algorithms-as-a-Service” with founders from Algorithmia.io, Wise.io, and the VP of Marketing from Revolution Analytics. As a relative newcomer to the world of data and algorithms, hearing the opinions of those who have spent years in this space provided a great learning opportunity.
One interesting topic of discussion was how to monetize algorithms as a service. As can be expected for such a new spacethe business model thesis is very much TBD. What I mean by this, is the big, industry-changing, IPO-justifying business model hasn’t been hit upon by the new class of companies in this space. Responses to the business model question ranged from revenue share from an algorithm marketplace, selling appliances that embed the algorithms for faster modeling, models built on specific machine learning algorithms, API use, and premium features. As far as customer acquisition – still very much hand-to-hand combat.
In my opinion, companies that are successful in this space will develop use cases on top of the algorithms. Use cases serve two specific functions – they answer the question “what problem does this solve?” and they enable the measurement of ROI. Without answering these questions the ability to migrate from a data science tool to a business solution will be limited – and the money lies in business solutions.