Developing an Acquisition Modeling Strategy
It's possible to make this useful tool even better. By combining two modeling techniques, you can vastly improve the quality of data from a single-model approach and increase the effectiveness of your marketing efforts.
There are two primary types of predictive models: response models and prospect value models. The former predict the likelihood that each prospect will respond to a marketing offer and the latter estimate each prospect's value to you once it is "acquired."
In a response model, each prospect is rated with a percentage likelihood to respond to your offer. For example, prospect A may have a 20 percent likelihood of responding while prospect B has a 5 percent likelihood. Obviously, you should solicit prospect A before spending money on prospect B.
A prospect value model assigns a dollar value to each acquired customer. So, if prospect A is worth $25 and prospect B is worth $100, you should solicit B before A.
Marketers who only use a response model can fall into the trap of ignoring the dollar value of a prospect. Combining the data of a response and prospect value model will let you build a more complete picture of your prospect pool.
Say, for instance, that you are only using a response model to target your prospects. Your response model has done an excellent job predicting the prospects that will respond to your offer and your response rates have increased. Your acquisition strategy looks great.
Then you (and your boss) find that the prospects you targeted with your response model were the least profitable ones. Your acquisition modeling strategy was unable to determine the value of your prospects. It's only a matter of time before you discover that these new responders cost more to prospect and retain than the dollars they will spend with you.
The same holds true for prospect value models. You might do a fine job of targeting the most profitable prospects, but the likelihood of acquiring them is so scant that they become unprofitable.
It is key to account for response and revenue by using both models to create a combined model score. This is done by building a response model based on responders and nonresponders from your previous acquisition mailings and a prospect value model based on the initial purchase amount from your previous responders.
Now, assign each prospect a score from each model, independent from one another. You will use these two figures to calculate a combined model score by multiplying them together.
Prospect Response Score Revenue Score Combined Score
A .20 $ 25 $5
B .05 $100 $5
As you can see, these two prospects look very different when their response model and prospect value model scores are viewed individually. But when both pieces of the puzzle are fit together, you see that their combined scores are equal.
So what does this mean? It's simple. The combined score is the estimated initial revenue you can expect to receive from each prospect. Say you solicited 100 prospects with a score of $5 each, you can expect to receive a total of $500 from your prospect universe. If this universe consisted entirely of prospect A's, you would expect to get 20 responders who each give $25. If the universe consisted only of prospect B's, you would expect to get five responders who each give $100.
This leads to another use for the combined score -- as an acquisition allowance. You interpret the combined score as the amount you can spend on each prospect to break even. In the example above, you can expect to spend up to $5 on each of your 100 prospects (whether they are A's or B's). It's important to note that these combined scores are based on initial purchase amount and not on long-term or lifetime value.
Used together, the response and prospect value model techniques become much more powerful database marketing tools than when they are used separately. By correctly implementing these two techniques, you'll be able to increase your revenue and response rates while also making more effective use of your marketing dollars by targeting the prospects with the most profitable combined score.
Tim Berry is vice president of analytical services at Merkle Direct Marketing Inc., Lanham, MD.