Before founding TROVE, a predictive data science technology company, I took two wireless data companies from inception to successful exit. I mention this, because wireless carriers wanted Platforms – mobile Ad platforms, messaging platforms, location platforms, you name it. If you wanted to solve a problem for wireless carriers at scale, you developed a platform. Period.
I brought this platform thinking with me into the predictive data science field with TROVE and thought for certain I was on the right track, because everything in the space seemed to be pitched as a platform war:
Whose platform could process the most data the most quickly?
Which platform had the most APIs?
Which company would emerge as the platform standard? Would it be IBM with Watson? GE with Predix? A specialty provider in its respective niche?
With the benefit of time, I can now safely say the answer was, and is, none of them.
Shakespeare wrote “The play’s the thing”; I’m here to tell you: “The platform’s not the thing.”
And I say that having built one, a great one in fact. Yet experience has taught me that enterprise clients don’t have time for data science platforms. They want answers not platforms. And they need those answers quickly, with demonstrable ROI.
By the time you develop and implement a platform, a more agile solution could have produced value across multiple use cases.
Customers also don’t have the faith – nor should they – that making data useful should require a multi-million-dollar gamble on a platform. “If we build it, they will come” worked great in Field of Dreams, but it’s not in sync with the business and predictive data science needs of today’s enterprise.
So, if not a platform, what is right for the enterprise? Agile Data Science.
Agile Data Science is a collaborative learning process that helps clients identify the best use cases for predictive data science, i.e., those that will deliver the most value. It then combines the great problem-solving abilities and modeling sophistication of data scientists with the elegance and convenience of configurable predictive software – leveraging AI, machine & deep learning, and other advanced techniques – to deliver answers fast.
Sound simple? Yes, in the way the best solutions usually do. But, remember, it took me years of selling a predictive data science platform to get there.
And that’s not to say there isn’t a role for platforms in this space. Of course there is. Platforms play a pivotal role in Agile Data Science, providing production scale. But the path to better data-driven operations doesn’t start with a platform. It can even be hindered by one.
Some of these ideas came into clearer focus for me recently as TROVE brought on a new Managing Director of Product Management, Tom Martin. He grappled with these issues firsthand at Pacific Gas & Electric while running their Emerging Grid Technologies Group. I want to share something with you that he told me that I think you will find useful:
“There has been a lot of money thrown at the idea of a giant, magical data platform that is somehow going to solve all of your problems, when what is really needed is an acute focus on the use cases of that data and those analytics, and how they will capture value for the end user, for the line of business, and for the customer.
“Instead of taking a top-down approach that starts with a platform and next turns to figuring out the use cases to layer on top of it, we need to reverse the order. Establish the use cases first. Identify the opportunities to take a more data-driven approach in the business to squeeze value out of the way it is currently operating, so it can be done more efficiently at greater cost savings to the business and its customers.”
Applying predictive analytics through an Agile Data Science framework puts the focus on these opportunities, providing the speed-to-value the enterprise needs to make the bigger case for prediction at scale – and the ability to operationalize it.