Case Study

Leveraging Credit Expertise to Forecast Customer Hardship and Minimize Bank Charge-Offs


Credit Risk
Modeling and Forecasting
Business Analytics
Business Process Automation


Kevin Baquero


The recoveries and customer support group at a Top 25 bank found itself facing a turbulent economic landscape in 2020 when the COVID-19 pandemic slowed the entire world’s economy to a crawl. 

Flying Phase created and then rapidly deployed a pandemic-sensitive call center capacity planning process that effectively forecasted delinquencies, supported staff increases of seventy-five percent and cleared governance and compliance approvals with zero findings or remediations. 


  • Effectively supported staffing increases of seven-five percent to address increases in delinquencies and customers facing hardship.
  • Cleared three separate governance and compliance approvals with zero findings or remediations
  • Enhanced portfolio monitoring to mid-month (on-demand), which allows forecasting to be finely turned during the month.
  • Supported the bank in recovering more than $45 million in payments in the second half of 2020


The recoveries and customer support group at a Top 25 bank found itself facing a turbulent economic landscape in 2020 when the COVID-19 pandemic slowed the entire world’s economy to a crawl. Like most banks, the organization relied on models and forecasting to anticipate customer delinquencies and hardship and efficiently staff its call centers to work with those customers to recover payments and bring accounts current.

COVID-19, with intermittent surges in cases sometimes followed by lockdowns and spikes in record-setting jobless claims, practically blinded these models. This was paired with unprecedented levels of government action, as well as uncertainty on the political landscape. No existing model could capture credit risk in this rapidly devolving economic scenario. And without adequate support, customers would likely default on loans en masse, which would adversely impact them and the bank.

The key to being able to right-size staff to effectively assist customers who had fallen behind in payments or who needed assistance with forbearance or deferral programs was precise forecasting and adaptability. But, given the magnitude of economic volatility and the speed with which it was occurring – not to mention the fact that any model would struggle in such uncertain times – the bank’s challenge became less about modeling and more about credit expertise in preparing for delinquencies, assisting those customers and limiting their losses in the progress.


Create and then rapidly deploy a pandemic-sensitive process that would enable the bank to forecast customer hardships and then plan staffing capacity at call centers to support those customers.

“The team focused on developing a lean forecasting solution that would be quick to stand-up in production and easy to execute.”


Instead of viewing the delinquency forecasting challenge in a vacuum, we began by evaluating the process in the context of how it affects bank operations and the customerexperience. We focused on how the bank’s operations organizations (the call centers) would use the process to actually help customers manage their debt. The goal was to build a new forecasting method that not only captured the impact of a rapidly changing economic environment on credit performance, but also considered how results were used downstream to best equip those decision makers to take action.

Through research and empathy sessions, we learned key nuances that would be integrated into the model. We particularly noted that not all forecasting errors were necessarily equal. The cost of credit losses from underpredicting delinquencies and then understaffing the call centers to help those customers was between five and ten times the cost of overpredicting call volumes and hiring excess staff. Additional considerations such as the type and complexity of training needed for each type of loan, customer and call center agent also factored into the new process.

Timing was another key focus in framing the problem. Operational leaders would need weeks to recruit and train new agents to accommodate understaffed call centers. Appropriately ramping down staffing when the peak had passed was also taken into account. The team focused on developing a lean forecasting solution that would be quick to stand-up in production and easy to execute. These features would give operations decision makers immediate visibility into their business and the flexibility to adjust call center capacity regularly as pandemic cases fluctuated, stimulus payments reached consumers and circumstances evolved.

Over the course of four weeks, we built a new forecasting solution based on current delinquency behavior (e.g. roll-rates, deferral status, balances, etc.), historical recession-era credit trends and pandemic-specific economic scenarios. We factored in seasonality considerations and also monitored the whole portfolio of credit and other accounts – including checking account balances and spending habits – to enable the bank to better anticipate customer hardship and delinquency before it happens. For example, if the bank knew that checking account balances were down and that customers were not cash liquid, they could anticipate an increase in delinquencies. Moreover, knowing that the economic outlook was deteriorating based on reports from Moody’s, the bank could blend employment figures and other outlooks into the forecast.

The new system incorporated qualitative factors such as Federal relief programs and bank deferral programs. We then segmented the forecast to consider different loan products separately and put additional monitoring in place for loans in forbearance and deferral programs. This focus allowed us to better prepare the bank for potential risk when those programs ended.

A modular, plug-and-play design allowed for rapid updates and refits as the bank’s staffing levels, call volumes, and other factors fluctuated. Integrating the forecast with the capacity planning process
empowered operations analysts to account for a variety of “What if?” scenarios and prepare for a range of potential outcomes based on, for instance, the timing of stimulus payments, new lockdowns, and the extension of relief programs. We then shepherded the entire solution through multiple layers of credit, model risk, fair lending, and legal reviews to ensure full compliance with bank policy. This approach ensured the effectiveness of the solution and provided the bank a compliant and well-managed solution for planning capacity throughout the pandemic.