The collection industry has the difficult job of contacting and collecting from debtors. How do you determine which accounts should be actively pursued? One common method is to use a collection scoring model.
Consumer debt collectors spend valuable time, money and resources to try to collect from debtors who aren’t going to pay or are difficult to reach. Predictive collection scoring models can help optimize the allocation of collection resources by identifying the most effective collection strategies. The end result would significantly increase collections and reduce collections costs.
A collection scoring model enables better collection decisions based on a multitude of available information. There are many vendors that have them available for your use. Typically, solution providers who offer scoring solutions include hundreds of variables in their models.
Basic Method of How to Create a Scoring Model
The basic requirement is to determine what information is available. Typically, this includes, but not limited to a person’s name, address, phone number, social security number, co-signer, account open date, account closed date, placement level, type of debt, balance, last payment date, place of employment (POE). You then place each variable into their own field in an excel spreadsheet. Once the fields are populated, you can add weights or values to the populated data fields. The weighting can start off subjective and can be refined after validation. Here are a few examples of weighting of the variables:
Variable 1
Are all fields populated (person’s name, address, phone number etc.)? Are they available to us all of the time? The answer would either be a 1 for YES or 0 for No. We place the response in each variable field.
Variable 2
For the “account open date” field, ask: can aging be determined from time of account open to today? If less than six months insert a 1. If less than a year insert .75, if less than two years insert .5, if less than three years insert .25. If older place a 0.
Variable 3
For the “account closed date” field, ask: can we set up aging from account closed to today? If less than six months insert a 1. If less than a year insert .75, if less than two years insert .5, if less than three years insert .25. If older place a 0.
Variable 4
For the “last payment date” field, ask: can we set up aging from last payment date to today (indicative of future payment)? If less than six months insert a 2. If less than a year insert 1.5, if less than 18 months insert 1, if less than two years insert .5 and if older place a 0.
You can continue to add variables to the spreadsheet that you deem important. Once all variables are in place, you add them up to create a predictive score. The score will then be utilized to prioritize your collections. [See table below.]
| Account Number | Name | Address | Zip Code | Phone | SSN | Co-Signer | Account Open Date | Account Closed Date | Placement Level | Type of Debt | Balance | Last Payment Date | POE | Score |
| 9998759 | 1.00 | 1.00 | 0.40 | 1.00 | 0.00 | 0.00 | 0.50 | 1.00 | 1.00 | 1.00 | 2.00 | 0.75 | 1.00 | 10.65 |
| 9998754 | 1.00 | 1.00 | 0.20 | 0.00 | 0.00 | 0.00 | 0.50 | 1.00 | 1.00 | 0.50 | 1.00 | 2.00 | 0.00 | 8.20 |
| 9998543 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 0.75 | 1.00 | 1.00 | 1.00 | 1.50 | 1.00 | 1.00 | 13.25 |
Collection strategies are now based on your predictive score. Treat the best 25% of the accounts with your strongest collection strategy and step down on efforts and costs on the next 25%. For the lowest group work only the minimum required.
After a month or so, you can validate the model to review performance. Do not make changes or modifications until the model runs a full collection cycle. After 180 days the “last payment date” variable may be a stronger weight. You might determine “co-debtor” isn’t as helpful as you originally thought. Statistical models are only as good as the data used to drive them. They need to be continuously validated as time and socioeconomic changes occur.
You now have built a predictive model based on your history and experience, which should enable you to find more likely payers and should provide you with greater cost efficiencies.
Professional scoring and analytic companies have more efficient and cost effective models readily available for your use, empowered by PHD-level statisticians.
Dwayne A. Banasiak, Vice President Consumer Collection Products at SunGard, offers knowledge in the application of analytics and decision sciences for collections.