CUSTOMER PREDICTIVE MODEL
Use Predictive Analytics to Create a
Customer Model to Target Your Best Audience
Customer Predictive Model – What Sets Us Apart
The Team at Inbound Insight has run many Analytic Models and Reports for clients. We know the best ways to ensure good results and will provide you with the benefit of our knowledge. We know when to use this type of analysis vs. Customer Profile Reports and Segment Conversion Reports. We can also advise you on the best ways to select the prospect records for postal, phone or email campaigns (e.g., in a region or multi-radius) and how to best onboard those records to maximize digital reach.
Key Benefits
The key benefit of the Analytic Model and accompanying report is that unlike Profile reports which “descriptive” in nature, the analytic model is “predictive” based on the creation of a custom regression model based on the data provided. This model scores every U.S. household from 1 to 100 with “1” being the most likely to have the characteristics of your customers.
In addition, the custom model is applied to all households in a top consumer database so new prospects can be ordered based on their score. So if you need to reach 30,000 new prospects in a region in which there are 300,000 total households, you could order all households that score in the top 10% or decile. (Other demographic filters can also be applied on top of the model score, as required.)
Using a multi-variate approach invariably provides better results than using the results of a descriptive profile report in which you manually choose continuous ranges from a few single dimensions (e.g., age 50 to 79, Income $100K+) to approximate the best results.
Customer Predictive Model – Background
Analytic Reports summarize the principal components and their values that are the most relevant to the predictive regression model. Analytic Model Component
We start by appending third-party data to the name and address data provided, then use statistical techniques to create a custom model. The results of this model are then used to score to a U.S. database of all adults on a scale of 1 to 100 where households that score a “1” are the most likely to be just like your customers.
Analytic models take as their input a customer file (5K to 50K records). By applying multi-variate regression, it’s possible to identity the 5 to 8 principal demographic components of the regression equation and their weights. For example, here are two of the seven principal components from a prior analysis.
Once a model has been developed, the decile scores are applied to the entire U.S. Household population. Those households predicted to be most likely to respond are in the top decile and would product the highest lift vs. a campaign sent out to random households. Here the lift associated with decile 1 is 3.65x. The lift associated with deciles 4 to 10 are below average.
To get a predictive look alike audience we run counts in the target area based on decile so we can determine how deep into the deciles to go to get the desired quantity. (Note: existing customers would be suppressed out. Other demographic filters can also be applied.) Once the decile and other specifications are known, we can get the targeted mailing, phone or email lists.
The results of this type of analytic approach are likely to exceed those of demographic profile method because the regression model uses many more variables to pick out and prioritize the best target audience.
Customer Predictive Model – Use Cases
Here are some key use cases for customer analytic models:
- The primary use case is to find a look alike audience that closely mimics a client’s current customer base. Usually this is in the form of a mailing list or email list.
- The secondary use case is to understand the most important demographic elements and values from the report that comes with the model.
Customer Analytic Model – Documentation
To download a sample Customer Profile Report, click the image to the right.
