Customer score calculation

The customer score evaluates and categorizes customer risk levels based on these five factors, calculated by Gen AI:

• Periods final score: The Periods Weighted Score represents the overall performance or status of customer accounts over five defined periods (Period1, Period2, Period3, Period4, and Period5). By using AI, the values from these periods are summed, normalized, and weighted according to Fibonacci numbers (1, 2, 3, 5, 8) assigned to each period. A final score is calculated that ranges from 0 to 160, with 0 indicates the highest risk and scores closer to 160 indicates lower risk.
• Credit used final score: The Credit used score represents how much credit of a customer has been used. By using AI, the credit utilization percentage is calculated with the formula: (Order Balance / Credit Limit) * 100. If the credit limit is zero, the credit utilization percentage is set to 0. If the credit limit is less than the order balance but not zero, the utilization percentage is capped at 100. The values for order balance, credit limit, and credit utilization percentage are normalized. These factors are weighted as follows:
  • 20% for order balance
  • 30% for credit limit
  • 50% for credit utilization percentage

  A final score is calculated on a scale from 0 to 160, where 0 indicates the highest risk and a score closer to 160 indicates the lowest risk.

• On hold final score: The On Hold score represents the total number of times a customer was placed on hold, based on data from the last five years. By using AI, you can analyze how frequently the customer appeared across all ARSC variants and tracked instances of being placed on hold. A hold was recorded whenever Order balance is more than credit limit. From these instances, you can calculate the percentage of times the customer was on hold relative to total appearances. This percentage was then scaled into a score ranging from 0 to 160, with a score of 160 assigned if the customer was never placed on hold (i.e., 0% on-hold percentage). As the frequency of being on hold increased, the score decreased accordingly.
• Payment pattern final score: The Payment Pattern Score represents payment behavior of a customer of the last five years, analyzing the delays between payment dates and due dates. Payments are classified into different categories based on their lateness that is, late by 3 months, 2 months, 1 month, 1 week and payments made within a week. By using AI, these categories are normalized and weighted with Fibonacci values (1, 2, 3, 5, 8) assigned to each period. A final score is calculated on a scale from 0 to 160, where 0 indicates the highest risk and scores closer to 160 indicates lower risk, suggesting timely payments.
• True payments final score: This score represents the total number of payments a customer has made on time of the last five years. By using AI, the total number of payments made by the customer was calculated and identified the number of late payments. To determine the True Payments, subtracted the count of late payments from the total payments. Calculate the average number of payments made by all customers. After normalizing the total number of payments, scores were assigned on a scale from 0 to 160. A score of 0 indicates that either no transactions are made, or all payments are late, while a score of 160 indicates that all payments are made on time.

The customer score is the sum of these factors, with a maximum of 800. The weightage for each factor is as follows:

Factor	Weight	Score range
Periods final score	30	0-160
Credit used final score	30	0-160
On hold final score	10	0-160
Payment pattern final score	15	0-160
True payments final score	15	0-160

Customers are categorized into high, medium, or low risk, where a score of 0 indicates high risk and 800 indicates low risk. The analysis is based on data from the past five years, with customers clustered using the K-means algorithm.