With all the big data hoopla, only one thing matters to a marketing leader: the relevant insight. Insight driven by predictive analytics is the basis for smarter decision making.

Predictive Analytics into the FutureFor a membership organization, for example, the insight (from an analysis of its data) may simply be that members who purchase a two year membership have a much higher retention rate than do those who sign up only for a year.

Voila! The organization begins to promote its two year memberships and sees its retention rate rise over the course of a couple years. Then the organization discovers that two year memberships are a predictor of those who will likely make an additional donation to the institution.

What are such insights worth? Priceless. That’s the power of predictive analytics.

The Power of Predictive Analytics and Controlling Data Gone Wild.

A few decades into the digital era, we now have the data and computing power to build sophisticated customer profiles. In February 2012, The New York Times published an article about how Target, the retailer, was able to predict when a woman was pregnant based on her buying behaviors.

Suddenly, predictive analytics hit the mainstream, and Eric Siegel, author of Predictive Analytics: The Power to Predict Who Will Click, Buy, Lie, or Die, became a poster boy for data science. A former professor at Columbia University, he was interviewed on Fox News and quoted widely as an analytics expert. Siegel is the founder of Predictive Analytics World and is Executive Editor of the Predictive Analytics Times.

The book Predictive Analytics is itself a rambling piece of unstructured data. The hardback runs 300 pages, and, frankly, its natural length is probably closer to half that. I recommend reading only the foreword, the introduction, and chapter seven (“Persuasion by the Numbers”). Siegel could have benefited from a stronger editorial hand in structuring his argument.

Still, the book gives marketers a good overview of the popular but technical subject. For example, Siegel clarifies some of the terms and science of the craft: “Although it’s sometimes called data mining, PA (predictive analytics) doesn’t ‘drill down’ to peer at any individual’s data. Instead, PA actually ‘rolls up,’ learning patterns that hold true in general by way of rote number crunching…”

He defines the task of machine learning as finding “patterns that appear not only in the data at hand but in general, so that what is learned will hold true in the new situations never yet encountered.”

In Chapter 7, Siegel addresses the difficulty of perceiving persuasion. Put simply, how can you tell from the data that a specific mailing, for example, actually persuaded someone to buy? Siegel says simply looking at the response rate is too crude: “The response rate completely overlooks how many would buy anyway, even if not contacted.”

He writes, “Standard response models predict the wrong thing and are in fact falsely named. Response models don’t predict the response caused by contact; they predict buying in light of the contact.”

Siegel advocates the “uplift modeling” approach: Will the customer buy only if contacted? Siegel then drills down into the specifics of uplift modeling, a bit too complex for most readers. His explanation of the simple problem, though, alerts marketers to be suspicious of simple response rate analysis.

But So What?

I’m weary of the “big data is here” wave of blogs, articles, and books that have pounded the marketing shoreline in recent years. It reminds me of the “social is here” tidal wave a few years before that. For marketers, the subject of data is existential: What are the insights from “my data” that can increase response rates (or retention rates) and lower costs?

After reading Predictive Analytics, I came away with two general observations: First, it’s important to keep the subject of data and predictive analytics in perspective.

My younger brother is an oncologist (a pharmacogenomist) who oversees clinical research and drug development at one of the world’s leading medical research centers. One assumption that he says guides his research is a fundamental distrust of data. Often it’s dirty, incomplete. How can you analyze a clinical trial of 2,500 women when you can’t verify whether every person actually took the drug every day of the trial?

My brother’s distrust of data makes him a better scientist, forcing him to be more cautious. That’s a wise perspective for marketers. It’s important that predictive analytics takes its rightful place in the pantheon of marketing disciplines. The tool, however, is just one of many to prepare people to buy.

Second, it’s now time for the analyst to have a seat at the strategic table.

In the old days, when you needed analysis, you had to ping the IT department, take a ticket, and wait a few days for the report. Then you’d wait a few more days if you wanted additional analysis. In the last 10 years, the BI (business intelligence) tools have matured to the point that the person who provides the initial cut at the analysis can answer some of your follow-up questions on the fly while you’re seated next to her in a meeting.

This dynamic interaction among team members is critical to the discovery of relevant insights and their application to the next campaign and making predictive analytics work.