With accounts grouped into clusters of similar customers, it stands to reason that each “in-group” customer should be able to spend in the range of the upper percentile of the cohort to which they belong (e.g., 80th). For the 20% of customers whose sales exceed the 80th percentile of sales for that group, their potential is stepped up incrementally until the cluster’s high water mark is hit. Here’s how it works:
Figure 2: Example Potential Table for a Single Cluster
For each of the individual clusters identified, there is a potential table created based on the revenue percentiles within that cluster. In the illustration above, each account within that cluster will receive one of four potential values. All accounts below the 85th percentile will receive that value for potential. Accounts between the 85th & 90th percentiles will receive the 90th percentile, accounts between the 90th & 95th percentiles will receive the 95th percentile value, and so on. Ultimately this results in a set of absolute potential metrics equal to the number of cut lines multiplied by the number of clusters. If, in this example, the company had 11 clusters the model would assign one of 44 potential values to every account.
Figure 3: Generating Unique Whitespace Values