Outside a train rumbles by

September 9, 2016

"The Fourth Industrial Revolution is being driven by a staggering range of new technologies that are blurring the boundaries between people, the internet and the physical world. It's a convergence of the digital, physical and biological spheres." --Fulvia Montresor, Director of the World Economic Forum, 2016

Did you notice whenever an L train (part of Chicago's public transit system) rumbles by our office, a little train rumbles by in the footer of our web page? Well, now you know: it's almost like you're in our office.

Why would we do such a thing? Mostly for fun, and we've got a soft spot for the train because every one of our four offices so far has been adjacent to a purple line train. But also, it's an example of one of the important changes happening that gets us excited about the future of technology and data science in particular: the blurring boundary between the digital and physical worlds.

The initial rise of data science was mostly driven by "digital footprints" -- basically, data created by people using digital things. Along the way, a lot of useful applications grew not just from log files and databases but "unstructured data" -- basically, data created by people writing things in natural language. Now, we're seeing another slew of projects not only from "digital footprints" and "unstructured data" but "sensor traces", basically -- data created by physical things such as cars, homes, and even ourselves.

We're excited about this blurring boundary because it opens up a universe of the most exciting type of data science opportunities: the type where there is an important problem, previously only approached with qualitative reasoning, that can suddenly be approached scientifically (quantitatively). But these opportunities with the IoT are deceptively ambiguous: the ability to execute often lags our imagination.

The general IoT strategy that we've seen so far in the wild is something like this:


It's an unfortunate reality that organizations spend too much time and money planning and developing IoT infrastructure only to realize retrospectively the opportunity they were pursuing wasn't well formulated.

So, what's a better way? Approach the problem like an entrepreneur or a designer. With ambiguous opportunities (like those that come along with the IoT), planning and infrastructure development are less important. Why invest in a plan that will be obsolete? Make iteration cycles fast and less expensive by starting with quick, low-fidelity prototypes while the opportunities are still ambiguous. Only transition to higher-fidelity prototypes and production infrastructure as you start to see where the real value lies. In our experience, it's the only way to avoid costly IoT mistakes.

contributors to this post

headshot of Mike Stringer
headshot of Jess Freaner