Continue in 2 seconds

So You Have a Model, Now What?

Published
  • February 21 2003, 1:00am EST

David Coppock would like to thank Drew Talbot, vice president and senior consultant of ANALYTICi, a division of IPG, for contributing this month’s column.

Imagine a data mining and modeling project that goes flawlessly. After some hard work to get the data, make it usable, select variables and write programs, the model validates beautifully. To the modeler, the findings look like they might be really important. A meeting is held, the findings are discussed and then…nothing. Weeks go by. The database has been scored and the findings discussed again with the project sponsors, the agency and various others, but no tests begin. The project quietly dies.

Why?

More often than not, the reason this happens has nothing to do with the skill of the modeler or the quality of data. Here are four reasons why a good model may not get the chance to drive a successful program.

  • The people in charge of the project did not formulate actionable insights.
  • The sponsors of the work did not communicate the insights derived to key constituents.
  • The results don’t agree with institutional truths.
  • The project never had a sponsor and champion.

Formulate Actionable Insights

Insights are actionable because they say something about behavior that can be controlled. For example, if a retailer finds out that the top 5 percent of customers drive 80 percent of revenue, that’s interesting, but so what? What merchandise should buyers order and when? What promotional offers does the marketing department create? Who gets the offers and how? What are sales associates supposed to do/say?

Only when the insights are framed by what people in the business do, and might do differently, will they be actionable. Suppose we know certain customers will buy more stuff next month but are much more likely to buy a big screen TV or anything else we sell if we first sell them a DVD player. Furthermore, they are likely to stop buying for a while when we sell them a big screen TV. Now we can make some decisions about sales priorities, sales scripts, product placement and promotional offers.

CEO’s may not care what customers think. On the other hand, they might care about how information can be used to influence purchase behavior. Whether or not a modeling project ever has the CEO’s ear, it won’t influence the behavior of the organization unless there are clear statements of what to do to drive sales and/or profits. If modeling projects are going to produce these kinds of statements, they can’t begin without business requirements definition. They can’t proceed without a clear understanding of who the key constituents are, what kind of system or process they operate in, what problems they face and what outcomes they can influence.

Insight formulation takes real time and real work. At the beginning of the project there must be time to interview constituents and agree on scope and project goals as well as review data structure. Interviewing and data analysis may or may not be the special skill of the modeler. The modeler needs to benefit from due diligence, not necessarily perform all the tasks. At the end of the project the modeler has to collaborate with others to draw clear and relevant conclusions.

Which brings us to communication.

Communicating the Insights

The insights can be actionable but not acted upon. Failure to act happens because the actors either don’t get it or don’t want to get it and aren’t held accountable. In the case of not getting it, if what they hear is statistical gobbledygook then they can be excused.

Communication does not mean a PowerPoint deck and a phone call. The insights need to be evangelized. As the proverbial preacher said, one must “first tell ‘em what I’ll tell ‘em, then I tells ‘em, then I tells ‘em what I told ‘em.” Constituents need to be sold, not just told. They need to get the razzle dazzle, not just the facts. This is true in every organization, about every message, every time. If the people who do the modeling can’t do this (they probably can’t) then they need to work with others who can make them understand the insights and send them to work. Someone who specializes in communications will be invaluable in translating the insight the modeler turns out into a story that a diverse group can understand and then act upon.

Agree with Institutional Truths

Agree with institutional truths – or be prepared to go to extremes.

If we tell the merchants who buy the stuff that goes in the stores that the consumers aren’t the people they think, they will not believe us. If the accepted wisdom is that the best customers are males between the ages of 14 – 30, but we believe they are really middle-age dads, we must show the proof. If the conclusions are fact based and well supported, there is a chance of winning converts; but this never happens without a struggle. It may seem safer to tell them their cherished beliefs are true and stop; but if so, what did they pay us for? Good analysis is often wasted because of lack of courage rather than lack of science.

No Sponsor? No Champion? No Action!

Don’t expect change unless the people in power want change. If the model shows that customer lifetime value will actually increase if the company mails fewer offers, absolutely nothing will happen until the marketing people who are tasked with product sales numbers on a daily basis are told what to stop doing by someone in authority.

Conclusion

The hard work of turning data into insights that are acted upon requires statistical and data mining smarts. No less important are leadership, communications skills and an understanding of the culture of the organization. A team that does this work effectively includes people with diverse backgrounds who can pull together all of the required skills. The successful team functions as a unit to develop insights and turn them into action.

Register or login for access to this item and much more

All Information Management content is archived after seven days.

Community members receive:
  • All recent and archived articles
  • Conference offers and updates
  • A full menu of enewsletter options
  • Web seminars, white papers, ebooks

Don't have an account? Register for Free Unlimited Access