Do Talk to Strangers: Basics of Statistical Prospecting Models

Statistical prospecting models don’t have to be confusing to direct mailers. If you know about two main types of acquisition models, you can boost your business far beyond the "mail-to-everyone-all-the-time" approach to finding new customers. You don’t even have to know how to create models. Just know what to ask for when you want them.

Prospecting models let you target your mailings on people who you don’t know yet. The two main types of models are called “lookalike” (or clone) models and response models. These two models are good starting points because acquiring new customers is much more expensive than retaining existing customers and every penny you can save on getting a new customer can be reinvested in keeping that customer longer.

The look-alike models are so named because they are built from the most important characteristics of your current customers and then applied by identifying the same characteristics to a file of prospects to pick out the ones that look like your best customers. You will need to have more than just the names and addresses and annual purchase of your current customers. Your file can be enhanced by appending dozens of demographics variables such as household income, family size, home value and the head of the household’s education level. Your analysts or marketing services vendor can use your customers’ average purchase amount and the demographics to create an equation or model.

This model can be applied to rental files which have also been enhanced with demographics to apply a score or rank to each prospect record. The highest scored prospects are the ones who look most like your best customers. When they are sorted from high to low and broken into 10% groups or deciles, you can choose just the top 10, 20 or 30% of the list to mail, confident that they are the ones most likely to buy from you over time.

The second kind of model is a response model. If you want to limit the number of mailings or keep a low frequency of mailings this model won’t tell you exactly who will become a customer, but instead who is most likely to respond to your mailing. This difference can be hard to understand, but is important.

The look-alike model predicts eventual customer acquisition, but not a reaction to a specific type of mailing. The response model predicts who will react to a mailing and not specifically who will become a higher value customer. If a phone call or response to your mailing is the most important criteria, then a response model is the right tool.

For a call response model you will need to mail your usual direct mail vehicle; letter, self-mailer or catalog to a sample of prospects. Track the mailing by later marking off or matching back the prospects who call your firm. Many firms equate marketing success with making the phones ring so a phone response model will help meet that requirement.

By using a process very similar to that used to develop a clone model, your analysts or marketing services firm can use appended demographics and the knowledge of which prospects responded and which ones did not to identify the most significant characteristics that separate callers from non-callers.

From this, a model equation will be created that can be applied to score new prospect records that you rent. When these prospect records are sorted from high to low, you will be able to select the top three or four deciles for mailing. Subsequently, you will know that you are targeting the most likely callers while saving a lot on printing and postage by not mailing the prospects who are not likely to respond.

There are more models that you can use to select specific segments, niches of prospects, or those who will appreciate special or seasonal offers. However, the concepts of prospecting by look-alike and by response models are best to understand first. Once you have these in your marketing toolkit you will be able to talk to strangers when you want to, knowing that you are talking to the ones who will want to hear what you have to say.

Bill Singleton is a manager of analytic services at The Allant Group in Naperville, IL.