Alan Kay, a fascinating computer scientist, famously said, “the best way to predict the future is to invent it.” He practiced what he preached. Among other things, in the early 1970s he invented eBooks and tablet computers—some 30 years before the Kindle! One can’t always invent the future (we’re not all Alan Kays), but we can often nudge it in the right direction. Kay’s central insight was that we should switch from being reactive to being proactive.
Prediction allows us to make this switch. Here’s a great example to help illustrate: The UCSF Hospital in San Francisco was having trouble with patients missing appointments. Calling the “no-shows” wasn’t helping fill the doctors’ slots. They built a model to predict who was likely to miss an appointment. Then, instead of using it simply to estimate how heavily to overbook, which is a weak form of being proactive, they looked more deeply at the model. Many of the patients that missed appointments lived three hours outside of San Francisco—a city with notoriously bad traffic. A better solution was to set up more local clinics so patients could be seen in their home community. And that’s what they did.
The Evolution of Data Science
Data science is shifting from descriptive and diagnostic (i.e., what happened and why did it happen?) to predictive and prescriptive (i.e., what is likely to happen and what should we do about it?). Insights from the predictive models can allow one to “invent the future.” This means you can respond to observed demand for a product or service by designing products and advertising to activate the latent demand. In other words, you can anticipate what people need—and bake it into your offering. A great example for the broadband industry might be a broadband service provider (BSP) offering security controls to subscribers before they ever ask. This addresses their need by activating it. Subscribers might not realize they need security controls until it’s too late. By then, the experience is stressful and poor—something we’d all like to avoid.
Leveraging the Power of Predictive Data
Likewise, responding to customer complaints about broadband service performance is worse than predicting that a customer is likely to be impacted by network performance issues. Anticipating future issues and proactively addressing them before they arise? Even better. This brings us to what broadband operations teams need most: good data. In order to act ahead of time, you must know ahead of time. Data is how you can accomplish this, and you need tools that help extract this data and analyze it so you can take action.
The next wave in data science will require better understanding of what predictive models tell you about subscribers, your network, and the services delivered over your network. Predictive analysis of trends in broadband operations can suggest where teams may need to develop new network design and operations strategies to improve the subscriber experience. What devices will subscribers connect? How much bandwidth will they consume and what applications will they use? What new managed services will emerge? Don’t just react to the changing world. Use predictive insights to invent your future.
Connect with Lyle Ungar, professor of computer and information science at the University of Pennsylvania, on LinkedIn.
Copyright © Lyle Ungar 2022