Stop the Insanity!

Most database marketers are fooling themselves when they think they understand what is driving their existing customers to buy from them. This is because they try to allocate responses to individual promotions, using source codes and matchback processing. As we will see shortly, this is a fool's mission.

Database marketers agree that, in today's world, the name of the game is multi-channel synergies. However, when they try to allocate each response to a single promotion, they are implicitly assuming that the response was caused solely by that one promotion. If this were the case, then why talk about multi-channel synergies?

Trying to allocate responses to individual promotions, using source codes and matchback processing, does not provide real insight into what it is that is actually driving customer behavior. There are two reasons why:

  • Promotional cannibalization. Source codes overstate the effect of some promotions, and understate others. For example, source codes typically overstate the effect of email, and often dramatically.
  • A significant amount of revenue would occur even in the absence of direct marketing promotions. Our firm refers to this as "Baseline Revenue," and it varies dramatically by customer. For certain types of customers, it can be as high as 80% or more. For others, it can be 20% or less.

Baseline Revenue is the natural result of the "drive-by" promotional value of an e-commerce site, which acts as a retail store. Other contributors include brand loyalty, and "background" promotional media such as paid and natural search, banner ads, and affiliate marketing.

The Insanity of Source Codes and Matchbacks

The allocation of response behavior to individual promotions is a daunting task, even if it made any sense. This is because of the massive amount of multi-channel promotional overlap that is so typical in today's world. For example, catalog/e-commerce companies often drop six or more mail pieces during the Christmas season, and supplement them with weekly emails. Then, layer on Internet sources such as search, and paid affiliates, and the complexity is staggering!

The Five-Steps to Sanity

The following is a five-step path to database marketing sanity:

Step #1: Focus your testing and measurement, as well as your segmentation and circulation management, on logical time periods such as seasons, and not on individual promotions.

Step #2: For each of your customer types, estimate the true incremental effects of promotions across multiple channels; that is, net of Baseline Revenue. This is done by analyzing your marketing database for past "natural tests," and by running going-forward longitudinal ("over time") tests.

We live in a world where, because of promotional cannibalization and Baseline Revenue, there is no guaranteed relationship between stimulus and response. In such a world, the response device means little, even when a source code is included. Therefore, we need to adopt an over-time ("longitudinal") input/output approach to testing and measurement, where different promotional mixes are compared by overall P&L.

Step #3: Re-think the role of Source Codes.

Intelligent source codes are an effective way to organize information. They are a short-hand descriptor for planning and tracking. However, they should not be used to make promotional decisions.

Step #4: Replace your existing segmentation approach with statistics-based predictive models that incorporate the knowledge gained from the previous three steps.

The predictions must span a logical period of time, such as the Holiday season. Also, for each customer, the dependent variable must be properly adjusted to take into account the number and "intensity" of the promotions received within the time period. Likewise, the dependent variable must exclude Baseline Revenue, which varies dramatically by customer.

Step #5: Continue to test, monitor and refine as necessary.

You must continually refine your estimates of the true incremental effects of your multi-channel promotional mix, within logical time periods. This is because the percentage of Baseline Revenue changes over time, especially as people become more and more comfortable with the Web.


When you follow these five steps, you will streamline your database marketing efforts. Even more importantly, you will begin to truly understand what it is that is driving the behavior of your existing customers.