# Enhance Collections Operations through Optimization

## Collections optimization can help organizations make better decisions by applying mathematics to complex problems.

Most of us understand the dictionary definition of ‘Optimization’: “the mathematical procedures of making something effective as possible.”  What most do not fully appreciate or haven’t seen in operation is how an optimization tool can truly enhance a collection program to increase revenue, make better use of staff, and simultaneously increase the customer experience.

Collection managers must constantly balance competing workloads and distribute their limited staff in order to find the best way to work accounts and maximize results.  Managers must determine which consumers to call, how often to call, and what treatment to select for those cases.  Historically, they use their experience to identify the optimal collections strategy.  Using technology, tools can help organizations calculate the best holistic strategy that best uses your staff and limited resources to increase collections.

Most large collection operations use predictive models.  Predictive models are mathematical algorithms to “score” cases for risk of non-payment, collectability, or likelihood to stick with a payment agreement.  These models are used to segment delinquent accounts so departments can either focus on the most collectable cases, or vary treatment to match the risk of the case.

The majority of models used by departments come from canned scores purchased from a third-party data provider.  More sophisticated departments take this a step further and build their own predictive models using historical data from their own operations.  Predictive models allow departments to better prioritize their inventory to determine which cases:

• Are the most collectable,
• Are likely to self-pay, allowing for lower cost, less intrusive treatments, and
• Are likely to only pay through stronger collection actions

While predictive models will improve your performance, these models alone can only increase performance so far.  For instance, predictive models do not prioritize cases on different workloads or help you calculate how many staff to assign to each workload.  That is where Collections Optimization (also called Prescriptive Models) comes in. Collection Optimization allows you to determine what is the maximum you can achieve with your limited staff and budget.

Optimization can help organizations make better decisions over management experience alone, by applying mathematics to this complex problem. Imagine a world where you had unlimited staff and the cost of collections was not an issue.  This would simplify the decision process.  You could give each agent just a handful of cases, and they could spend all day long vigorously pursuing those cases.  But that is not the world we live in.  The cost of collections matters.  Customer service matters.  Optimization therefore can help you maximize your results with your constrained staff level and following policies and statutes by analyzing the myriad of choices you have to determine which one brings the best result.

#### Insights from Optimization

Collections optimization delivers more than a strategy—it also provides insights based on multiple what-if scenarios that balance tradeoffs between goals and constraints, maximizing results with your finite staff.  Through optimization you can:

• Simulate strategy impacts by predicting reactions to different treatments.
• Determine what channels to use when contacting consumers (e.g., phone, SMS, letter)
• Select the best resources to assign to a particular case (e.g., senior collector, junior collector, or automated treatment), focusing your best resources on the cases that truly need their attention.
• Determine the timing for intervention, balancing your limited resources to know when to contact and how often to contact
• Drive better agency performance through Placement Optimization – give cases to the agency most likely to collect

Once consumer reactions to different actions are modeled, decision optimization commences. At this stage, the algorithms simultaneously apply all possible actions against constraints to determine outcomes. This step allows users to stress-test results and determine which strategies should be candidates for scenario analysis.

In scenario analysis, the tool can calculate the optimal operating point across multiple strategies. For example, the analytics may determine that one scenario allows you to retain the same level of collections with a 5% decrease of cost, while a second scenario increases collections by 10%, maintaining your baseline staffing levels.

By running multiple scenarios with different constraints, an efficient frontier is created. This efficient frontier sheds light on the range of possible outcomes, allowing managers to make an informed decision on the preferred operating point.

#### The Need for Automation in Optimization

Optimization is a complex mathematical problem, that is why you need a strong tool to build your models.  There are virtually an unlimited number of scenarios that could be evaluated.  In addition, there are any number of ways you can distribute staff, cases, collection strategies, etc.  For each possible strategy there are multiple metrics to be evaluated.  The optimization tool analyzes these large quantities of data, matches them to key constraints, and evaluates and scores alternatives to find the optimal strategy and distribution of resources.

Optimization can also show the best contact frequency and which cases can be collected with a lower cost contact method (e.g., SMS or email).  Optimization can also allow you to simultaneously reduce the number of aggressive collection treatments while still maintaining collection levels.  For example, you can find cases where you can avoid credit reporting, liens and garnishments, and instead collect through lower cost/touch approaches.

This will allow you to drive substantial gains, improving customer service by collecting with the lowest level of intrusion needed, lowering the cost of collections.

Have feedback on this blog post? Want to keep the conversation going? Connect with me on LinkedIn: https://www.linkedin.com/in/tedlondon/