Technology today affords us the ability to automate much of what was manual work ten years ago.  Businesses have implemented collection tools, cash application solutions, and dispute management modules which all increase productivity and efficiency for the company as a whole.  Credit management automation should be no exception.  A good start in making strides in credit management automation is through the creation of unique scoring models.

Scoring models are used in every aspect of our personal lives.  Our FICO (personal credit score) number is more important than our social security number today.  It determines our livelihood.  It’s our reason for being.  It determines whether we can get a replacement car for the old clunker that finally clunked three days ago or if we’ll be working at a fast food company for the rest of our lives instead of a Fortune 100 company.  The fact is, we are all products of a scoring model.  Scoring has ventured into commercial lending; not just on a consumer level.  Many businesses are now realizing the importance of risk management which soars above inconsistent, manual decision making.

Scoring models are so effective in risk management because they promote standardization when making credit decisions.  We can unknowingly play favorites with our customers.  A credit analyst could be more “easily persuaded” to make a bad decision because of a strong-arm customer on the other line, or any other reason which could deter proper lending decisions.  With scoring models, the risk assessment is consistent, which creates accountability for those times that a decision is made outside of the recommendation.

Having accounts assigned a risk score will quickly adapt meaning over time within the department.   The automation through models can increase productivity by scoring monthly accounts up for review and extending the review dates based on the model’s recommendation.  This automation allows analysts to spend more time reviewing larger and/or more risky accounts. Scoring will also enable the ability to perform focus reviews for possible credit limit increases on a large portfolio of accounts.

Want to develop your own scoring model?  It is feasible.  The process will be arduous and it takes dedicated people to pull it off, but I highly recommend making the effort to create an internal scoring model.  It is a good exercise to see what factors analysts within the department are using.  The development will also shed light on best practices and areas for improvement.

The first step to creating your own scoring model is to compile a Scoring Model Team that will take on the task.  The members should have the ability to devote at least five hours per week to the assignment.  The team should consist of at least two members with different skill sets.  Knowledge of Excel is a must, as the scoring model will be developed with Excel formulas and tested through Excel as well.  Expect this process to take a few months with brainstorming, development, testing and realization.

The best way to approach the brainstorming phase is to determine the factors used when making the credit decisions manually.  Is there more leverage given to a customer with twenty years’ experience compared to a customer with two years of payment performance?  What’s more important; their pay day performance or their financial Z-Score?  Surveying multiple users would be the best approach to determine what factors are consistently used. I recommend sticking with a list of no more than ten attributes.

Example Factors:

  • Years in Business
  • Days Beyond Terms
  • Net Worth
  • Security on File
  • FICO score

Once the factors are identified, development can begin.  The most difficult part is determining what type of weights to give each factor.  Years of manual credit analysis “thinking” says that every factor is equally important.  It will take time to break through this thought process to move forward to a model with selective weights based on statistical and mathematical observations.   This takes the most time but, creating the building block in Excel will allow for the ability to adjust the weights until you arrive at a formula that fits.  Each factor should total to 100%.

Example Factors with weights:

  • Years in Business  (25%)
  • Days Beyond Terms (25%)
  • Net Worth (15%)
  • FICO Score (25%)
  • Security on File (10%)

After determining the weights of each factor, a breakdown will need to be completed.  Each will have values/ranges associated with the possible values belonging to the particular factor. Below lists the “Years in Business” factor which determines in days or years how many points to assign a customer.  Logically, more points should be given to a long standing customer compared to a fairly new customer.

Table:

  • Years in Business
    • 0-90days  0pts
    • 90-1yr      5pts
    • 2-5yrs      10pts
    • >6yrs       15pts

Testing will quickly overlap with the development phase as the weights and points will be adjusted as the portfolio is scored and analyzed.  Testing should be used on 10% of the active customers.  The ultimate goal is to arrive at a risk score that would be a recommendation indicative of a manual decision.   Standard Distribution models will be helpful during this testing phase.

Once the model is completed, the model can be implemented for real life scenarios.  It is best to start on a smaller scale and utilize the new tool to make routine review decisions before venturing out to full automation.  A contributing factor in full automation is change management, which will take time as confidence is built amongst the team.

Also, see one of HighRadius’ previous posts, “The Transformation of Accounts Receivable – Characters Welcome!” to see how to identify various “characters” in an organization as it starts to transform their receivables management system.