Propensity to respond model

A Propensity to Response model is the theoretical probability that a sampled person (or unit) will become a respondent in an offer or survey. They are especially useful in the marketing field. A response likelihood model can have substantial cost savings as it can lead to lower mailing costs by identifying patrons who are very unlikely to respond to a particular offer. After segmenting these people out, the casino can then focus on only those most likely to take up the offer. A casino can identify the likelihood of response from all eligible patrons. After that, it can identify the most valuable patrons that are most likely to respond. This allows the casino to estimate the expected response from the most valuable patrons and eliminate mailing(s) to the patrons that are of lower worth and/or are unlikely to respond.

 

Occasionally, response likelihood models will lead to easy decisions, such as cutting out low worth patrons with a low likelihood of responding. However, more complex situations might arise since response models are never perfect. It doesn’t matter how good a model is or how accurate the historical data is, there is always a chance that a patron identified as unlikely to respond will respond. Thus, when making a decision about patrons identified as unlikely to respond to an offer, it is also important to balance that likelihood of response with the potential return on response.

A propensity to respond model would be built using historical information around marketing campaigns and it looks at predicting the likelihood a customer will respond to a marketing communication. The advantage of this model is that it strengthens the marketing strategy even more, beyond purely segmenting the customer base. It can further allow for improved ROI on the marketing budget, by identifying the likely number of respondents to be returned by a campaign.

Often a business’ marketing department will have an expected number of respondents or an expected response rate. By identifying those who are most likely to respond, the chances of meeting that expected number or rate of response is greatly improved. Gone are the days of marketing to an entire customer base. This is an unnecessary waste of the marketing budget and also runs the risk of annoying customers by touching them too often or with the wrong offer.

Again, a predictive model could be built which identifies those most likely to respond through to those least likely to respond. This would be done using customer metrics and historical campaign/marketing information that identifies those who responded and those who didn’t. Variables that have a significant association with the customer action are extracted and these form part of the prediction algorithm. Every customer is then given a score according to how likely they are to respond to a marketing campaign.