What Is Predictive Modeling
Predictive modeling uses statistical analysis of your existing customer data to create a mathematical profile of your ideal prospect. This profile is then applied to a large prospect database, scoring every record by how closely it resembles your best customers. The highest-scoring prospects are selected for your mailing — they are your look-alike audience.
The result is a mailing list that combines the breadth of a large compiled database with targeting precision that approaches or exceeds response list performance.
How the Process Works
The modeling process follows these steps: First, your customer file is analyzed — hundreds of demographic, behavioral, and geographic data points are appended to each record. Statistical algorithms identify which data points most strongly distinguish your customers from the general population.
These distinguishing characteristics are combined into a scoring model. The model is then applied to a large prospect universe (often millions of records), assigning each prospect a score. You mail the highest-scoring records and skip the rest.
The scoring is typically delivered in deciles — your prospects ranked from top 10% (most likely to respond) to bottom 10% (least likely). Testing across deciles reveals where to draw the line between profitable and unprofitable segments.
Types of Models
Response models predict who will respond to your specific offer, built from previous campaign data. Look-alike models find prospects who resemble your customer base, useful when campaign response data isn’t available. Revenue models predict not just who will respond but how much they’ll spend. Retention models predict which acquired customers will remain active long-term.
When Modeling Makes Sense
Predictive modeling is most valuable when you have a customer file of 10,000+ records to analyze, when you’re mailing into a large prospect universe where untargeted response rates are low, and when the cost of your mailing is high enough that improved targeting delivers meaningful savings. The upfront cost of model development ($5,000-$25,000 depending on complexity) is typically recouped within one or two campaigns through improved efficiency.
Getting Started
Predictive modeling starts with quality data — both your customer file and the prospect universe it’s applied to. Browse our list categories for prospect databases that can be scored and optimized through modeling, or contact us to discuss how predictive modeling can improve your campaign targeting.