Don't Let Measurement Take a Back Seat

Over the past few years, the popularity and use of analytics in database and direct marketing has grown significantly. This can be attributed to several factors including the growing awareness of the role that analytics can play in helping companies realize their business goals. Many companies that practice database and direct marketing are generally using a healthy and useful mix of the different types of analysis, but often fall short in the area of measurement.

Broadly speaking, there are four categories or types of analysis: identification, prediction, segmentation and measurement. Measurement refers to tracking and evaluating the performance of marketing programs and analytic models supporting those programs. Generally, measurement focuses on response rates, comparative profiles of responders and nonresponders and financial performance and ROI.

Marketers that do not measure adequately are doing themselves and their companies a disservice by not providing comprehensive information on which to base marketing and financial decisions. In some cases, marketers can put their own job or budget at risk when they can not demonstrate to senior management the accurate ROI of their programs or learn from those efforts.

Use control groups. You can provide management the information they need for decision making through planning and the incorporation of a few basic steps in the design and execution of database and direct marketing programs. One of the most obvious steps involves creating control groups for purchase behavior measurement. Once the households or businesses that will be targeted in a program are chosen, adequately sized control groups should be randomly selected and withheld from the marketing. This will allow for the measurement of the natural behavior that would occur in the absence of the marketing program and provide the basis for the calculation of the incremental lift in results that can be attributed to the program.

In some instances, this type of control group may not be feasible or not required because a product or service is only marketed through direct channels where no natural purchasing can be expected.

Determine what will be measured. You need to determine precisely what is being measured and for how long. If the focus of a direct mail campaign is to spur sales of product X, will only product X sales will be tracked or will other purchase behavior that could be influenced — either positively or negatively — also be measured?

For example, say your company is using a direct mail campaign to increase sales of a specific product. As a result of the campaign, incremental sales of that product increased by 20 percent. But simply to say that the campaign was a success because sales increased 20 percent would be misleading because the campaign may have indirectly influenced the sale of other products. These additional sales need to be part of the equation when measuring the full impact of the campaign.

You also need to determine whether the effects on broader and longer-term measures such as customer attrition will be monitored. Such factors may be critical since they could ultimately impact ROI calculations and judgments about the relative success of the marketing effort.

Ability to measure tied to data capture. The ability to measure properly relies on good data capture. In the example of product X, you would need to have the mail house provide a promotion file net of undeliverable mail. In addition, keycodes would be set up that identify the marketing treatment received by each individual targeted in the promotion.

Records in the promotional file should have a unique key that can be used to match back to the marketing database or operational files that will contain the behavioral information that will be examined. Likewise, if a predictive model was built to support the program, can the scored population be matched to the promotional file and the behavioral information to evaluate results by model decile? All of the fixed and variable cost information associated with the marketing program needs to be available so that ROI can be determined.

Furthermore, a central repository of measurement results should be created to facilitate continuous learning. The repository should include not only the parameters and results of program rollouts but also the results of testing that preceded the rollouts.

If each of these considerations is properly addressed, you will truly be in a position to prove the level of success or failure of your efforts and also achieve critical learning to improve future campaigns. And like other business investments, management will have accurate information required to evaluate the return on the marketing investments and be able to make more informed decisions about the level of those investments in the future.

John Young is a senior director in the Analytic Consulting Group at Epsilon, Burlington, MA. His e-mail address is [email protected]

Related Posts