Thomas Goetz
If you don’t build a capability early in your startup, you’re unlikely to get to it later
The reality of big data —
IN 2013, WHEN MY CO-FOUNDER AND I started Iodine, we—like pretty much any startup—suffered from certain delusions of grandeur. Our slick slide deck touted our esteemed pedigrees, our unfair advantages, and our uniquely brilliant business idea. We were talented, experienced, and absolutely original.
In truth, we were riding in on a pretty crowded bandwagon—one called big data. We, like a lot of others, saw an opportunity to gather large amounts of information (in our case, about people’s real-world experiences with medications) and put computation to work. Data would go into our black box, analytics would happen, and out would pop insights and predictions that correlated to better outcomes. Lives would be saved. Money would be made. Particularly for health care, data is widely seen as an elixir that will revolutionize an ossified industry, rooting out waste and failure and bringing intelligence to the marketplace. But the same big-data demos were being done in many sectors, including transpor- tation, security, agriculture, and finance. Angel List tracks nearly 5,000 big-data startups in its database.
The trouble is that many of these companies got distracted, you know, starting up a company, so the data and analytics projects were back-burnered, and then allowed to go cold. “Big-data confessional” is one of my favorite startup parlor games—getting other founders to admit they never got around to creating the intelligence engines they had once dreamed of. Time and again, short-term demands superseded long-term data dreams. I’ve come to believe that the only way to do big data right is to do it from day one. At Iodine, we had a long list of data- driven ambitions, many of which never saw the light of day. But we did, thankfully, build out some analytics capacity in our early days, and since then we have often gone back and improved or fixed our data warehouse and infrastructure. It helps that my co-founder’s pedigree is indeed exceptional. He spent time building data tools at Google. But our success here owes as much to just doing the work when there have been so many competing priorities.
And there’s the rub: As easy as it may be to talk big about big data, the actual doing is a slog— plumbing and janitorial tasks that take a lot of effort and yield marginal benefits. For many companies, that work seems less rewarding than doing the stuff that actually might drive hockey stick growth. Which is pretty rational, when you think about it. These days, big data doesn’t quite have the buzz it once did, but its progeny—machine learning and deep learning and artificial intelligence—are sown into pitch decks and business plans like magic beans. The same story applies: If you’re planning on using one of them someday, then it’s probably never going to happen.
And this goes beyond data and analytics. Any startup with some sliver of change-theworld ambition has a two-step strategy. Today we’re doing X, but tomorrow we’ll be doing Y—that’s what we’re really building, they whisper. Trouble is, tomorrow is a long way off. And if X starts to drive revenue and growth, it’s almost impossible to shift to Y without jeopardizing any current success.
If you really want to be driven by data or A.I. or deep learning or whatever, the best time to create that capacity is on day one. Figure out how it aligns with your business plan and revenue strategy. And if it doesn’t, either change the business or stop kidding yourself that you’ll get around to it. The best time to build the startup of your dreams is before you’ve started building anything else.