By: Bart Baesens, Seppe vanden Broucke
This QA first appeared in Data Science Briefings, the DataMiningApps newsletter as a “Free Tweet Consulting Experience” — where we answer a data science or analytics question of 140 characters maximum. Also want to submit your question? Just Tweet us @DataMiningApps. Want to remain anonymous? Then send us a direct message and we’ll keep all your details private. Subscribe now for free if you want to be the first to receive our articles and stay up to data on data science news, or follow us @DataMiningApps.
You asked: What is response modeling and how can analytics be used for it?
Our answer:
When you are running a marketing campaign, it is not always possible or even desirable to target your entire customer base. A first obvious reason is limited marketing budgets: sending out irrelevant campaigns to uninterested customers is a waste of money. Moreover, some customers might even get so annoyed that it creates an adverse effect towards your product, brand or company. Hence, it would be ideal if we could send out the marketing message only to those customers who really benefit from it. The analytical solution to this is called response modeling. The goal hereby is to develop a classification model (e.g., logistic regression or decision tree) that selects the customers who are most likely to respond and take action. Put differently, response modeling focuses on deepening or recovering customer relationships using analytically based models.
Various types of responses can be considered. Assume we invested in an email marketing message or a Facebook Ad which has a fancy title together with a link to your website. The response can now be qualified in various ways. More specifically, we can make a distinction between a soft response and a hard response. Seeing the ad could be a first type of response that is of interest since it will create product and/or brand awareness. A next type of response would be someone clicking the link. This already illustrates more engagement with your offer. Responses may then come in a variety of forms. Examples are opening a web page or pdf with a product description, or leaving your contact details together with a request for a price quote. These are all examples of soft responses with increasing interest in your message and offer. The ultimate, hard response would be a product or service purchase. Hence, as a first step in response modeling, the data scientist needs to discuss with the business expert what target (e.g., ad impression, click link, pdf download, or actual purchase) needs to be modeled.
Once the response target has been appropriately defined, the historical data for analytical modeling need to be gathered from previous marketing campaigns in order to properly understand customer response behavior. Popular examples of data that might be useful are:
- Demographic variables (e.g., age, gender, marital status, employment status)
- Relationship variables (e.g., length of relationship, number of products purchased)
- RFM variables (see above).
- Social network information (e.g., purchase behavior of friends, product reviews from friends)
All of the above data elements can then be gathered into a dataset to build your analytical response model. Given the multitude of predictors available, it will be important to perform variable selection to make the model compact and powerful.
Various analytical techniques can be used for response modeling such as logistic regression, decision trees, neural networks and random forests. In fact, many companies feel comfortable using black box analytical models (e.g., random forests, neural networks) for response modeling since their primary goal is to find out who will respond rather than understand why customers respond. Note that besides classification, response modeling can also be approached from a regression angle, whereby the aim is to build a regression model predicting the amount (or intensity) of the response.
