How to go from vertical lists to broad market

Share this content:
Alex Aigner
Alex Aigner

With the average cost of a vertical list around $125/M, marketers can cut their data cost almost in half by switching to a national demographic file supplemented by a predictive model. Vertical lists by default are specialized groupings of individuals rather than mass millions. Utilizing one data source with the thousands of attributes it contains eliminates the time is takes to manage several lists and controls the cost of your merge/ purge efforts.

With a potential universe of approximately 200 million households, a successful broad market campaign must have an effective predictive model to identify the households that will meet the marketer's response, sales, and ultimately profitability requirements.

The unique challenge is that the direct mail campaign experience (based on vertical lists) is very dissimilar to the population that the model will be deployed on (using a national file). Observed trends and relationships that are present in the vertically restricted training population and which drive the discrimination of the model may not be present in the demographically diverse population. Without controlling the inherent biases between the training and application populations, such models will provide unpredictable results that will not meet expectations.

Segmentation of the broad market universe to determine the level of similarity to the vertical list experience is an essential feature of a successful broad market rollout. Segments that carry a very similar feature set to that of the past vertical list experience will exhibit the highest levels of accuracy when a traditional response or sales model is applied. Such a model is developed in a similar fashion to a profile model. A representative sample of the broad market population is pulled and serves as the modeling domain, all variables in the compiled list serve as the independent variables, and all records in the sample that exist in the mail experience are flagged and serve as the dependent variable for this exercise. The marketer should also include a random population, which will help expand the marketable universe.

Alex Aigner is EVP of business development at Datalab. Reach him at


You have until Wednesday, December 7 to get your entries in. Learn more here.


Next Article in Datamine