Situation
Client was looking to develop a single-sourced multi-channel database.

Objective
To cost effectively develop a unified database platform for postal and email marketing efforts.

Methodology
• Analyze each data source
• Create database design through
  procedures
• Improve creative to be concise and
  response driven
• Normalize and process data hygiene
• Develop database categories,
  masterfiles and enhancement files
• Combine multi-source data such as
  magazine subscriptions, email
  newsletters, event attendees, and
  product purchasers

Results
Revenue increased, complaints were virtually eliminated, deliverability increased significantly, and tracking and profiling of respondents was implemented.

Home > Database Services Overview > Database Design > Merge Purge > List Maintenance & Fulfillment > Analysis & Modeling

ANALYSIS & MODELING

First, we get to know your business and understand your objectives. Then, we conduct an analysis with an experienced understanding of direct marketing dynamics from your viewpoint. Employing the latest methods of regression and multi-variate analysis, we develop profiles of your customer. To help you improve response and better interpret campaign data, we've developed many market analysis programs that will meet your unique analysis criteria.

Applied Info Group's experienced programmers and statisticians work toward presenting pertinent, actionable, and measurable results. Our Good Customer Match and Mailed Regression programs help you arrive at extremely predictive algorithms. Ultimately, you will see lifts in your response rates.

Modeling Methodology

Our approach to modeling is identifying the highest quality prospects in the most efficient and cost-effective manner, while gaining invaluable information about your customer file. In brief, we follow a four-step plan:


Data Preparation and Enhancement
Pre-process files to standardize name and address information to improve match rates. Enhance with data from Applied Info Group Consumer file at household level. Perform preliminary data analysis including correlation, distribution, step-wise, and cross-tabulation to identify possible predictor variables.

Regression Modeling
Apply logistic regression model techniques to build a regression scoring model with the calibration sample. Then we generate your response gains table.

Model Performance and Validation
Validate the performance of the model on a fresh sample.

Regression Result Presentation
Results are presented and described in detail. Easy to read reports feature scoring and ranking of the prospect file and profiling for each group of prospect's names based on certain characteristics. Then, recommendations are made as to which duo-decile (5%) groups should be tested.