We receive marketing messages generated by predictive analytics daily. You’re ready to check out on Amazon.com when you are queued to a screen that says, “Customers who bought ‘The Da Vinci Code’ also purchased ‘Secrets of the Code.’ “
Retailers often use predictive analytics to offer cross-sells and upsells to customers based on past purchases. According to IDC, predictive analytic projects yield a median ROI of 145 percent. This is welcome news for publishers who are eager to shore up subscriber bases to preserve subscription and ad revenue.
The Seattle Times Co. uses database and advanced statistical technology to marry attitudinal survey data from subscribers and non-subscribers, along with marketwide geo-demographic data. This lets the publisher identify which news and information readership segment each subscriber falls into. That, in turn, lets it predict readership segments for each non-subscriber in the metro area, then test direct marketing strategies for increased retention, subscriber acquisition and chained discount programs.
The newspaper has perfected its approach over time, starting with limited, off-the-shelf software. The breakthrough came when it moved to a predictive analytic solution that provided an accurate way to integrate its marketing segmentation models with its marketing database. With this approach, the Times harnesses in-house marketing intelligence and segments at the individual level, targeting multiple messages within a household. For example, the Times can be confident that in a particular household of three the father likely falls into the “hard news” segment; the mother into the “avid loyalist” segment and their teenage son into the “wired” segment. But how do you get there? Here are six steps:
Secure commitment across the organization. Key departments – marketing research, circulation, IT – must collaborate to ensure consistency and quality while gathering data, interpreting models, executing campaigns and integrating predictive models with the marketing database.
Build a database. Now you need a place to store and tag your subscriber and non-subscriber data. A simple marketing database able to refresh the predictive model as new individuals are scored is essential.
Evaluate the accuracy of third-party data appends. To predict non-subscriber interests you need geo-demographic data for your marketing area. Virtually every third-party data supplier promises the highest level of accuracy, so test before you buy. Request a sample file you can use to test vendor claims – well worth the extra time.
Schedule iterative model reviews. Iterative model reviews are crucial. Too often, analysis projects are a recipe for unused shelfware: Your in-house or outsourced tech guru is dispatched to build a complex statistical model and deliver it “complete” to the marketing team. Marketers should be in at the start to define the business questions to be answered, contribute business logic and help determine what variables should drive the model for the most useful predictions. Weekly or biweekly model reviews keep everyone focused and the project on track.
Test the model. Before you go further, test the model on a sample from your marketing database. Testing lets you evaluate the model’s performance and identify ways to tweak it for greater accuracy before going live with your entire database.
Automate model updating. Through the marketing database, instruct the predictive model to deploy on a pre-established schedule. The benefits? Minimal maintenance, regularly refreshed behavioral data and updated predictors so that each time you target new prospects, you take advantage of the most recent data and best intelligence.
Predictive modeling gives publishers a way to use what they know, leveraging data assets already collected, identify new high-potential subscribers and increase revenue. Ironically, it was the technology advancements of the late 1990s that lured subscribers away from print newspapers and magazines, and the same refined technology now can bring them back.
