The value of predictive analytics seems obvious: who wants to “drive looking out of the rear view mirror”?
Several recent books, articles and blog postings point to a resurgence of interest in the topic, but what the term actually means is — as usual in our industry — subject to some debate. I’ve tended to use the term to refer to predictive models, overlapping with the term “data mining”, a sentiment echoed by David Loshin in this post. Professor Ian Ayres has collected a fun selection of simple predictive models on his web site, for everything from predicting life expectancy to the success of a book title to.
Despite the obvious uses of this type of predictive analytics in organizations (here’s a 2004 article that outlines marketing applications, for example), it has not been implemented widely. There are many reasons for this, including lack of BI maturity, the need for deep expertise, distrust of “black box” solutions that can’t “explain” the prediction, etc.
But perhaps the biggest reason is that people simply don’t seem to think it works in real life: simple models are too simplistic to be used outside of vendor demos, and even the most sophisticated models and technology soon break down in today’s fast-changing businesses. The cost and effort of implementing something that would actually be useful seem to outweigh the possible gains — especially because business people aren’t necessarily ready to believe the predictions!
Another type of predictive analytics probably has a rosier future. James Taylor defines the term more widely, encompassing “a variety of mathematical techniques that derive insight from data with one clear-cut goal: find the best action for a given situation” including “analytic disciplines used to improve customer decisions” and lays out his point of view on how it relates to BI and data mining.
As usual, I have to partly disagree: BI has always been “actionable” — otherwise nobody would ever have spent money implementing it — and I personally view traditional BI and predictive analytics as different levels of sophistication, rather than being fundamentally different concepts.
Here’s an example of how this kind of predictive analytics can help with “next best action” marketing and Seth Grimes believes that it’s going to be next on the shopping list of existing BI players:
“So what are the next targets for the analytics companies? Predictive analytics…”
Well, what do you predict?
Comments
4 responses to “Who Really Wants Predictive Analytics?”
A Predictive Models is made up of a number of predictors, which are variable factors that are likely to influence future behavior or results. In marketing, for example, a customer’s gender, age, and purchase history might predict the likelihood of a future sale.
[…] Originally Posted by Timo Elliot on LINK […]
Hi, I am from Cequity solutions. We are a Analytical Marketing company. We help our clients to make the most out of their customer data and help them create better marketing strategies. Take a look at our blog http://blog.cequitysolutions.com/ for more insights on customer and marketing analytics.
So who wants predictive analytics? Because the very notion of analyzing one’s way to a “correct” prediction is so provocative, it’s hard to even make a comment that easily stays on point instead of swinging freely between paranoia and irrational exuberance… But let’s look at the enthusiasms this way: *everyone* wants predictive analytics; a subset of everyone wants to pay for it; a subset of that payers’ group is willing to trust it; and a subset of that trusting group is willing to depend on it. And the difference between what makes predictive analytics successfully useful versus successfully marketable is a gap just as large as the miniscule percent of dependents versus the whopping percentage of enthusiasts. To figure out where you ought to be within that gap, you have to decide whether you need to bet on change or not, to get what you insist on gaining. The more you need to bet on change, the more you’re ready to rely on predictive analytics. “Predictive analytics” is confusing to people because it makes them think of “calculation” which makes them think of “facts” — and in other words encourages the idea that there’s science leading to certainty. The problem is that it isn’t the certainty they suspect.The real point of prediction is not to identify future outcomes; it is to identify future possibilities.