Model Types and When to Use Them
Targeting is arguably the most important variable in contributing to direct marketing success.
Taking it a step further, I would argue that the phrase “junk mail” is often a reaction to consumers’ perception that they got a piece of mail not meant for them. That makes direct marketers cringe. We don’t want to spend money reaching out to prospects who have no interest in what we’re selling.
The key to efficient targeting is a good mailing list, and that’s where modeling plays a role.
The theory behind modeling is that past success is the best predictor of future opportunity.
There are two broad modeling avenues that direct marketers can use to generate targeted lists:
The theory is simple. Look at your existing customers and find people just like them. Better yet, look at your best customers and find people just like them. Here’s how it works:
- The client provides current customer data
- Demographics are appended to the data
- A multivariate regression model is applied to that customer demographic data to build a model
- That model is applied to a prospect file
- Prospects are broken out into deciles based on the strength of the match
- Prospect lists are generated based on the most influential attributes
The strength of the cloning model is that it is relatively quick to generate from an existing customer file of just a few thousand names.
The cautions are that it cannot account for a customer’s buying preferences in terms of channel, offer or other variables related specifically to direct marketing. Another consideration is that cloning models work for what and how you market today. Current product and marketing strategy will dictate model results.
In this case, we don’t just look at your customer list, we look at who responds to your direct marketing efforts. Why is that important? Let’s say your business is built on a media mix of radio, search and display advertising. We can model your customers, but that doesn’t tell us their propensity to respond to direct mail.
Response models solve for that by analyzing the most important attributes of those who respond to your direct marketing effort. It works like this:
- Direct marketing efforts are tracked to determine who responds and converts. These can be broken out by offers or other appeals.
- The most significant attributes of responders and non-responders are identified
- Multivariate regression models determine responder file demographics
- The file is scored and broken into deciles ranking the file on likelihood to respond
- A prospect list is then scored based on those rankings
The strength of this model is obvious. The modeling is based on the most predictive criteria available for direct marketing success.
The downside is that it requires response data, and that can be a challenge for clients with smaller mailing quantities. It also requires tracking that gets back to individual mailing records, such as a match-back or other direct marketing customer attribution model.
This is meant as a general overview on modeling. Your marketing specifics and data will determine how modeling is executed for your company. Thanks to our partner Speedeon Data for this infographic that provides a quick overview of many of these points.
Infographic: Are you choosing the right data model?