Custom Response Modeling

Response modeling is used to enhance the performance of direct marketing campaigns.  Past campaign results are used to forecast future performance as it relates to demographic, socioeconomic, or other available characteristics. Information about promotions and the respondents' purchase history are statistically compared using multiple regression techniques in order to develop a predictive profile.  

Defining the Response Measure:  Determining and defining an appropriate measure of ‘response' is vital to the modeling process. DSRA works closely with our clients to determine exactly how best to measure the response characteristics which matter to a particular business. This response measure can range from simple responsiveness, to a measure of long-term purchase history or value, and may also incorporate such characteristics as bad debt or product returns.

Developing a Rich Predictor Set:  It is also important to develop a rich set of possible predictive attributes for the modeling process. Such attributes can include information about prospective customers such as zip code demographics, sourcing information, and characteristics determined from the name and address such as gender, urban/rural designation, or single home vs. apartment. If promotions are based on in-house prospect lists, additional attributes might be generated from the information held in the file. 

Creating the Customized Model: To create a custom model, the client will deliver information about a previous campaign to DSRA. This could be a House File extract or ZIP Code summaries of mail, response, and revenue performance. We can even factor in shipping expense, product cost, dunning cost and consider the multiple response patterns in continuation campaigns. DSRA will then create a custom model that segments high productivity candidates from low productivity candidates. The response modeling techniques determine which attributes affect the defined response outcome measure, and create a formula to predict ‘responsiveness'. Each prospective customer is scored according to responsiveness based on their unique set of attributes. Prospects can then be ranked by the predicted response score, and promotional campaigns targeted to reach the most likely responders.