Customer Analytics > Optimizing Your Collection Efforts Part 3


Share on LinkedIn

In a previous post I laid out the process to create the collections database and initial collections analytics during acquisition. This post will focus on how to utilize the database and analytics within proactive communications in order to optimize collection efforts then the ongoing tracking and enhancements required.

1. Treatment Strategies: Analytics alone will not enhance collections efforts. The predictive analytics and segmentation systems must be converted into actionable treatment strategies which will 1)reduce the cost of collections and 2)ultimately, minimize the outstanding debt through increased collection payments. Treatment examples which need to be setup and tied to each individual collections customers include (again, these are examples)

  • Those customers with a low probability of paying and a low to medium outstanding balance may be placed aside and not proactively contacted as the collections cost could be greater than the predicted payment amount. Also, a certain percentage of collections customers will pay on their own regardless.
  • Customers who have a medium to high probability of paying and their receivable age is low (time since entering collections is low) should be aggressively pursued to pay ASAP as once that receivable ages the probability of payment significantly decreases.

When developing the initial collections treatments start simple (7 to 10 treatments should do) as you will always have time to enhance and add onto the treatments over time. More importantly, starting with a good number of distinct and actionable treatments, that everyone agrees to, will allow for quick learnings and organizational buy-in.

2. Tracking:

  • Once the initial treatments are setup enhance the database with current and historical treatments so they may be used in the customer communications.
  • Create random control groups, which are small samples of collections customers who receive a standard treatment or no proactive communications at all. These control groups allow for tracking of the true incremental impact of the collection treatment efforts and provide a means to optimize the predictive analytics through an un-contacted or unbiased group of collections customers.
  • Add the treatments and model scores to the contact tables in the database
  • Track the results of the treatments (no answer, answer but did not pay, payment received,..) in the database that may be tied directly to the contact history table at an individual contact level.
  • Create reports which show both the INCREMENTAL impact of the treatments as well as collections performance overall in a trending fashion.

3. Ongoing Enhancements: Ongoing enhancements to the system should include-

  • Continually enhance and refresh the models as most environments are dynamic which causes the performance of the models to erode over time
  • Continually test new treatments and compare them to control group performance to determine which treatments to keep and which to enhance or throw away altogether.
  • Once a collection effort contact and response history is available create response models which predict the probability of a customer paying given a proactive contact. These models, if used properly, will greatly enhance collections efforts as certain customers will pay regardless of being contacted while other customers must be contacted to push them to pay.

In summary, the key to an optimized collections process are predictive models, actionable treatments and an ongoing test-and-learn process. Customer should be assigned to treatments based on probability of payment and predicted payment amount (the models), outstanding balance, previous collection actions (ie. 2 outbound phone calls the past week) as well as other customer segments that will help tailor the collections efforts based on custom environments.

When developing a system from the ground-up it may be best to initially focus on cost cutting (minimize collections costs) while building the treatment performance history. Once enough performance history is available then work on maximizing the amount collected.

Republished with author's permission from original post.

Roman Lenzen
Roman Lenzen, Partner and Chief Data Scientist at Optumine, has delivered value added analytical processes to several industries for 20+ years. His significant analytical, technical, and business process experience provides a unique perspective on improving process efficiency and customer profitability. Roman was previously VP of Analytics at Quaero and Director of Analytics at Merkle. Roman's education includes a Bachelor of Science degree in Mathematics from Marquette University and Masters of Science in Statistics from DePaul University.


Please use comments to add value to the discussion. Maximum one link to an educational blog post or article. We will NOT PUBLISH brief comments like "good post," comments that mainly promote links, or comments with links to companies, products, or services.

Please enter your comment!
Please enter your name here