One of the primary functions of direct marketing, in addition to generating response and profits, is to continually learn what’s working, what’s not working and how to capitalize on that knowledge.
While most savvy marketers employ tests to help identify new ways to improve direct marketing performance, the testing methods commonly used are often expensive, cumbersome and do not allow for long-term optimization of response and profit. However, by employing some new modeling and test-design approaches, direct marketers can execute more efficient and more effective campaigns.
Problems of Inefficient Testing
Inefficient testing can cause trouble in a number of ways:
• You implement dozens of direct marketing tests but are not sure whether you are realizing the maximum benefit from your investments.
• You often don’t have answers to management questions because you haven’t actually tested the scenario they’d like to know about.
• If your campaigns tend to be small, you feel there is a limit to how much testing you can do and still yield statistically reliable results.
• You’d like to build more powerful response and profit models, but feel you’ve gotten about all you can out of your data and your modeling techniques.
Suppose a marketer uses three scores to segment a direct mail population: a response score, a risk score and a revenue score. Each score is analyzed in 10-point score bands, which yields a 1,000 cell matrix (10-by-10-by-10), assuming the marketer is planning just a single offer.
Using today’s techniques, generating enough observable and statistically reliable data – depending on expected response rates and loss percentages – might require mailing approximately 50,000 pieces per cell. Mailing 1,000 cells would require 50 million pieces. As you can see, doing even a few tests would exceed most direct marketing budgets.
What we typically do is compromise. Either we use fewer scores, we scale down the depth of our analysis – perhaps looking at 30-point score bands instead of 10-point score bands – or we collapse our analysis across various offers and marketing tactics, thereby not getting a true read on each variation. And, in most cases, we truncate the mailing, often selecting the number of cells we will mail.
Strategy Optimization Process
Getting beyond the current testing paradigm can be done by employing some new modeling and test-design approaches. Drawing from statistical techniques and scientific methods used in other industries such as engineering, manufacturing and medicine, there is a process – called the strategy optimization process – for developing new products and improving existing ones for direct marketers, enabling faster, cheaper and more effective direct marketing campaigns. The critical steps in this approach are:
• Develop predictive response and profit models that will quantify the effects of your marketing actions on consumer behavior. This is done by including characteristics of the offer design, the channel strategy and the marketing tactics within your predictive models. This will help you see whether there is a material impact on response and profit for each of your marketing actions and exactly how much each action will influence response and profit within each market segment.
• Using commercially available software, design a direct marketing mail plan – in consideration of your budget and other business constraints – to determine which cells and what quantities you will need to market to statistically infer results for all possible cells in the marketing matrix. Here you can begin to reduce the complexity of your campaigns by requiring fewer cells to be mailed and allowing for smaller sample sizes, while still generating the intelligence you would have gained from executing all the possible offers and strategies.
• Execute the direct marketing campaign ensuring you capture all the data elements required for back-end tracking and model development. While this step is not new, it’s critical to your ongoing success.
• Using the data obtained from your marketing campaign, infer results for all the possible cells in the mail matrix. Then, using results from the full mail matrix, develop the predictive response and profit models you specified to quantify the effects of your offer design, channel strategy and marketing tactics on consumer behavior. Calculate response rates and profitability for each possible cell in the matrix and evaluate results.
While this new approach may seem a bit complex and will require the guidance of experienced modelers who understand your business, you will soon realize how much faster you can learn what’s working and what’s not working within a far greater sea of possibilities. And you will more likely find those diamonds in the rough.
Lauren Jackson is product line manager for Fair, Isaac and Co. Inc.’s Targeting and Prospecting Solutions, San Rafael, CA. Her e-mail address is [email protected].
