Use Case: Targeted customer offerings

cancel
Showing results for 
Search instead for 
Did you mean: 

Use Case: Targeted customer offerings

Predict what is the next best offer to market to a telco customer


What’s the problem?

Companies have an increasingly large range of product offerings, resulting in a growing pool of customers. As the matrix of applicable product-to-user options expands, how does a company ensure they are marketing the right product to the right customer and maximizing their return on investment? They are facing a variety of significant issues: traditional methods of mass marketing are no longer as impactful, regulators are stepping in to fight spam, and customers are becoming more self-sufficient researchers rather than consuming the traditional marketing mediums. The result is a need to reinvent the way companies market to customers.

The challenge and solution

The answer is DataRobot’s multiclass modeling. Multiclass classification problems answer questions that have more than two possible outcomes or classes; for example, a question asking which of five competitors a customer turn to instead of simply asking whether or not the customer is likely to make a purchase. With additional class options, you can ask more “which one” questions, resulting in more nuanced models and solutions. So, instead of flooding customers with a wide range of products, DataRobot allows companies to evaluate and choose products best-suited to the individual. The benefit is two-fold: customers receive information only on potentially interesting products, and companies receive higher customer loyalty and increased marketing ROI by providing granular, hyper-targeted messages and offers.

For example…

Caesar is a marketing manager at a national telecommunications provider. He has been tasked with cross-selling existing clients on different plan offerings or features in order to hit this quarter’s revenue targets. Caesar’s job is to decide which plans and add-ons to market to specific customers. He collects all available plans plus upgrade options and an initial list of several thousand customers. With so many customers and several available plans that could be offered to each, it is not possible for him to manually select the optimal product to offer. While he lacks the technical ability to construct a highly sophisticated machine learning model himself, he can turn to DataRobot’s automated machine learning platform to build models that will predict which offer each respective customer is most likely to respond to.


Training data: https://s3.amazonaws.com/datarobot-use-case-datasets/DR_Demo_Telecomms_Churn_Multiclass_test.csv 3

Prediction data: https://s3.amazonaws.com/datarobot-use-case-datasets/DR_Demo_Telecomms_Churn_Multiclass_test.csv 3

Data Dictionary: https://s3.amazonaws.com/datarobot-use-case-datasets/Telco+Data+Dictionary.pdf

(Originally posted May 2018)

Version history
Revision #:
10 of 10
Last update:
‎12-17-2019 02:44 PM
Updated by: