- Thomas Thurston
Demystifying Data Science
To explore this a bit further, we spoke with Thomas Thurston, CEO of Growth Science, and the organizer of Innovator’s Anonymous, a unique workshop where innovators from the world’s top, most sophisticated businesses come together to dip their toes into the pool of data science in an unconventional way.
Mike: What’s the first thing you’d like people to know about predictive analytics and data science?
Thomas: I’d say, it’s all a lot more accessible than most people think. Anyone can start doing it. When you hear about people using algorithms and technology to crunch data and predict the future, it sounds like science fiction. 99% of the world thinks to themselves “that sounds really hard, I could never do that.” The truth is, believe it or not, building predictive models is something anyone can do.
Mike: What do you mean? What do people need if they want to start?
Thomas: They need a worthy question. The hard part is less about how to compute, and more about what to compute. Analytics are almost always built by technologists – people who really understand computing, algorithms, statistics. While there are definitely people doing interesting work in those fields, for the most part, right now I see a lot of technology looking for applications. A lot of hammers going around looking for nails. What’s lagging are the right business insights – the right questions – to put those tools to work for.
Mike: Are you saying the technology doesn’t matter?
Thomas: The technology definitely matters, but what I’m saying is, the tactical mechanics are far less important than the question itself. Mechanics should be an afterthought, not a gatekeeper. It’s too easy to get rat-holed on the technology, but none of that matters unless you know what question you’re trying to answer in the first place. What’s the strategic question? It’s been my experience that finding the right question can be harder than finding the right mechanics to answer it. One of our guiding principles is to never do analytics for the sake of itself. Everything has to be in service of the bigger question.
Mike: So let’s say you have a good question. Now what?
Thomas: Next you need a framework to take apart the question and build a model around it. By “framework” I don’t necessarily mean a technological framework. You need a way to think about your question and translate it into a predictive model.
Mike: Is this what people do in Innovators Anonymous?
Thomas: Yes, Innovators Anonymous is a very small, elite event where innovators from the world’s top firms get together and collaborate around a framework to turn questions into models.
Mike: Are these “compute” models you’re building at Innovators Anonymous or something else?
Thomas: When I say “framework” I’m talking about a way of thinking; a step-by-step way to translate a question into a model. Innovators Anonymous is about sharing our favorite framework and doing so in an unorthodox way. Sometimes when building a model you realize you’ll need a lot of compute horsepower, but a lot of times you don’t. The goal is to keep the mechanics as simple as possible and only use the bare minimum to get the job done. Sometimes there are hardly any statistics involved and if you’re lucky you can just Google what you need.
Mike: Are you saying you sometimes do your statistics using Google?
Thomas: Definitely, if we can. There are all kinds of free tools online. Sometimes you need bigger iron, but you’d be amazed how far you can get with a laptop and a little Googling. Again – it’s about using the least mechanics possible to get the job done. Granted, you have to be robust and know what the rules are, but beyond that it doesn’t have to be rocket science.
Mike: Lets talk more about Growth Science and the predictive simulations you do there. What can you tell us about that?
Thomas: We do what’s called predictive business model simulation. In short, we simulate business models in dynamic market environments to predict if they’ll survive or fail, among other things.
Mike: That sounds a lot more complicated than what we’ve been discussing so far.
Thomas: Well, yes and no. On the one hand, there’s no question our predictive tools have come a long way and now involve a relatively high degree of sophistication. But the important thing to remember is, we didn’t start there. We started with a few good questions like “will a business survive or fail” and went from there. Everything else came later. It’s a lot easier to start simple and then add-on over time, and that’s exactly what we did. If we’d tried to boil the ocean at first we’d probably still be sailing around in circles.
Mike: Any parting thoughts? What would you say to someone who wants to start exploring data science today?
Thomas: Just this, don’t be intimidated and don’t give up. There are some things our brains are incredible at, but there are other things our brains are notoriously bad at. Data science is just a new name for the very old idea of using data and frameworks when our brains need a little help. Now there’s more data than ever and every week there’s another fancy way to analyze it, but that’s a good thing. It’s getting easier and easier to do things that used to be really hard even a decade ago. These days nobody should consider predictive modeling beyond their grasp. With a little creativity and the right people to help you from time to time, you’d be amazed how far you can go. Everything’s in front of us. It’s an amazing time to be alive.
Mike Bernhardt, Publisher, The Exascale Report