How Do You Curb Bad Debt? Real-Time Analytics Can Help You Gauge Credit-Worthiness Without Angering Sensitive Customers


Share on LinkedIn

The trick to every business is managing real expenses while improving the customer experience. I have found that as technology has made it possible to manage analytics closer and closer to real time, it is helping businesses connect the two.

One recent client I worked with had a typical problem, even though it was not the average business setting. Our company was hired to help the prison communications system get a handle on the ongoing troublesome bad debt. It was a sensitive matter because the telephone is one of the few ways for inmates and their families to keep in touch.

The system could not be too stringent or revenue would be suppressed and the customers would be routed to customer service too often.

A primary provider of calling solutions to correctional facilities nationwide had a not-so-surprising problem: The bad debt rate was high. The write-offs were such a large concern that the CFO was the one who initially spearheaded the effort to tackle this issue.

The prisoners place calls to an outside line, and the calls are billed collect to those recipients. Initially, in an attempt to minimize bad debt, an imposed credit limit was placed on each call. The core issue was that the credit limits had very little intelligence behind them, so a number of customers were receiving too high a limit, causing the bad debt issue. Others were receiving too low of a limit, resulting in revenue suppression.

Further, the lack of intelligence behind the credit limit caused unneeded routing of consumers to customer service, which is costly to the client, and resulted in an unhappy customer experience for many. The client needed a solution that would optimize the credit limit based on true credit worthiness, thus optimizing revenue (not just bad debt) while enhancing the customer experience.

Buried within were many additional business issues and hurdles, including these:

  • Names and addresses were not available for many of the people the inmates were phoning, which made it difficult to profile those that were responsible for paying the bill.
  • Most calls were billed by the local exchange carriers (LECs: Quest, PacBell, Verizon), which meant that there was little visibility into whether an actual payment had been made, with such details as an individual call payment date and payment amount.
  • Each interaction included two customers, the prisoner and the phone recipient, who both needed to be satisfied. The prisoners needed the phone system to keep in touch with family, so they had a voice into its usefulness. The call recipient was responsible for paying the bill. This is a competitive industry, so it’s important to keep both customers happy. Their combined voice will be heard by the ultimate customer, the correctional facility.
  • The availability of historical data varied greatly. While a number of customers had several previous calls and payment transactions, which could be used to generate future payment predictors, other customers were first-time callers.

I needed to build a real-time analytic system that maximized revenue and was sensitive to a good customer experience. The system could not be too stringent or revenue would be suppressed and the customers would be routed to customer service too often. On the other hand, the credit limits could not be too loose or bad debt would remain at the same levels or even increase.

We started with several business and data assessment sessions with the client to discover and address all the issues and hurdles. A few of the hurdles could be directly addressed, while others required proxies. For example, company executives decided to build a multi-tier analytic system with several models depending on the number of previous calls (historical behavior) a customer may have had. At the same time, we identified outside data sources for identifying call recipients.

After months of development the analytic consultant team and internal client finance department delivered a series of models and financial solutions that were combined to properly assigned credit limits, based on previous call behavioral history and credit worthiness, using both internal and external data. Not surprisingly, we found that one of the most powerful drivers of future payment was previous call and payment behavior. Once the credit models were complete, a historical sample of customers were scored and compared to previous applied credit limits to determine the optimum future credit limit that should be applied to optimize revenue.

Our analytic and technical consultants then worked with the client’s internal IT team to develop the real-time scoring and credit-limit application module.

The initial results of our solution to this challenging problem show more than a 1,000 percent return on investment. In many business situations, driving incremental revenue is in sharp contrast to the customer experience, but it was not the case with our solution. Assigning proper credit limits increased the customer experience. Those people who were credit worthy were no longer prematurely cut off from their active calls and routed to customer service.

The application of real-time analytics is typically a journey, so we have identified many additional areas to improve results over time. As the company continues to enhance the solution, customer service will increase and bad debt will continue to decrease, resulting in ongoing incremental revenue and satisfied customers.

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