The Mother of All Models

No, we’re not talking about a high-cheekboned, exceptionally fertile female.

What I’m referring to is a tendency I see quite often in corporate America to “crack the code” on marketing measurement by building the world’s biggest regression model.

While admirable in their pursuit, companies that seek to answer the question “What are we getting in return for our marketing investment?” with a number, i.e., “41% ROI,” are headed into a long, dark alley with a penlight. Their chances of getting to a numerical answer with any high degree of confidence is about the same as their chances of finding a specific grain of sand on the ground in that alley.

Along the way, wonderful things are learned about correlations between marketing stimulus programs and business outcomes, most of it negative. In fact, the real value that model-seekers derive is a much higher level of clarity on what doesn’t work. So there is some value to pursuing it. But outside of some packaged goods categories with clearly defined and mature purchase patterns and competitive environments, this approach rarely results in anything close to a perfect prediction of economic outcome in relationship to changes in marketing investments.

The real question is, what is the real question? You see, if you’re trying to ascertain the optimal level of advertising spend to maximize either short-term profitability or ROI, you certainly can build some effective analytical approaches to get to a reasonably small range of uncertainty (a.k.a. high degree of confidence) in an answer. Marketing-mix models can be quite helpful in answering this and similar questions.

But if you’re trying to answer the more common CEO question, “What would happen if I spent twice as much on marketing or half as much?” the answer tends to elude the power of pure analytics absent years of detailed transactional data and previously determined influences of external variables like interest rates, housing starts, demographic mobility, etc.

The bigger the question, the more likely you are to need to a comprehensive combination of marketing metrics to assemble the preponderance of evidence, like a marketing dashboard. Using insights you gain from analytics plus test/control experiments plus research plus some structured forecasting techniques, a marketing dashboard helps focus the company on what it knows and, by definition, what it doesn’t know. Over time, the key questions are identified, researched, and answered.

In short, if your inclination is to try to tackle broad questions of marketing measurement through advanced modeling techniques, you’ve only got part of the solution. Using the full set of tools at your disposal to complement your analytics will enhance your overall ability to answer the really hard measurement questions with greater credibility.

If you’ve had any experience with a modeling-centric approach to marketing measurement, please share it with the rest of us.