Don’t Just Predict the Future, Learn to Change It!

If you think predictive analytics is of interest only to data scientists and “quants,” think again. It is fast becoming one of the most talked about topics both in business and business schools.

Two recent examples make excellent cases-in-point. I teach a class in new media marketing as an executive-in-residence in Cornell’s Johnson Graduate School of Management. I love giving back to my alma mater and am always intrigued with what’s top-of-mind for future marketing and entrepreneurial leaders. Guess what produced the biggest buzz in the classes I just taught? You got it – predictive analytics. And at the eMetrics Summit in San Francisco in mid-April, one of the week’s most exciting presentations was by our client, Mario Pacini, general manager and CMO of HP Snapfish Americas. Yes, the very same topic.

Of course, human beings have been trying to predict the future since antiquity. Here’s what media luminary Marshall McLuhan said about the problem: “We’re driving faster and faster into the future, trying to steer by using only the rear-view mirror.”

So, is predictive analytics about to fundamentally change how we market? Why the excitement? And what about that rear-view mirror?

The first point I make with my students and clients is that it’s not just about predicting the future. Leveraging sophisticated forecasting and optimization models, marketers now possess the much-coveted power to change the future.

General managers and CMOs have always wanted to know how to invest their money to achieve revenue and margin targets. Predictive analytics can help them determine whether they are investing funds in ways likely to meet top-line and bottom-line goals. And the data can also show them if current plans will result in a miss. They know before the fact so they can course-correct to avoid what every CMO or general manager fears – failing to meet revenue and profitability goals. Using predictive analytics, marketing organizations can identify the likely outcome of actions to support decision-making. And they can iteratively adjust those decisions to have increased confidence in outcomes.

How does predictive analytics work? How can businesses forecast with a high degree of accuracy and then determine what to do differently to achieve more desirable outcomes?

The entire field of analytics, of course, rests on a foundation of data. Massive amounts of data provide extraordinary diversity and depth of information about consumer purchase behaviors, interactions on websites, in social media and on mobile platforms, and, of course, in stores. We can now determine which campaigns and promotional offers, delivered over which channels to what devices, drive purchase behavior for specific buyer segments.

So, what can we do with all this big data other than use it as a rear-view mirror reflecting only the past?

This is where data scientists come in to do their magic. Statistical models are built and algorithms are applied to these large, multi-channel datasets to deliver predictions given a proposed course of action. You don’t like the forecast? Then apply optimization models to see different variations of the future that may be more appealing.

For a concrete illustration of this process, let’s review the example from the recent eMetrics Summit: HP Snapfish, the online photo service giant with more than 90 million members in more than 20 countries.

The HP Snapfish team has a great deal of data about what visitors do on the photo website. For example, who buys which products during the holiday season compared to the summer? Which customers are motivated by promotions, discounts, or delivery options? What drives customer loyalty and repeat purchase behaviors?

Predictive analytics involves deconstructing these levers to assess and identify historical performance drivers and then reconstructing them accordingly to drive future goals. HP Snapfish indicated that it breaks down its customers into different segments according to behavior, such as:

  • Purchases based on promotions/discounts vs. not
  • Types of products purchased
  • Purchase times of the year for different product categories
  • New vs. repeat customers

Based on patterns revealed, the company put together a promotion calendar that matches buyer behaviors. It discovered how one lever is related to another, and how they affect the outcome. It is, after all, about developing a model that is likely to deliver the goals needed, assess performance regularly, readjust the levers, and repeat. It’s a logical process.

About those rear-view mirrors?

Are we looking in the mirror and seeing only the past? In one way, we are. This is historical data. But in another way, we aren’t. These are sophisticated models that project the future with a lot of accuracy by identifying the factors that lead to an outcome. It’s not really a rear-view mirror, but a form of detection, if you will, in the complex arena of analytics.

But don’t forget my main point here.

Predictive analytics and optimization models turn all this data into truly actionable intelligence. They tell you what to do. So it really becomes about changing the future. That’s why data scientist has been called the sexiest job of the 21st century. They get to do what millennia of fortunetellers have failed to do. Who said destiny was predetermined anyway?

Image on home page via Shutterstock.

Related reading