Revenue flow
Gross sales from identified customers
Gross sales from identified customers
Ranked by net merchandise sales
| PRODUCT | NET SALES | UNITS SOLD | ORDERS | CREDIT / GROSS | SHARE OF SALES |
|---|
Identified customers only · full available purchase history
Recency × frequency · bubble size = historic spend
Historic gross sales by customer segment
Monthly repeat activity by first-purchase cohort · last 12 cohorts
What else lands in the basket, strongest links first
Baskets with both products · at least 5× chance
| METHOD | TYPICAL WEEKLY ERROR | 13-WEEK TOTAL ERROR |
|---|
Ranking quality across thresholds
What your decision threshold actually means
Find repeat buyers without flagging everyone
Held-out permutation importance · ROC AUC decrease
ROC AUC on the held-out snapshot · 0.5 is a coin toss
Average spend per tenth of customers, ranked by prediction
Predicted vs actual spend · log scale
Model vs ranking by past spend alone
Held-out permutation importance · R² decrease
Five behaviours, reduced to two dimensions
Separation and stability for 2 to 8 groups
Medians per group
| GROUP | CUSTOMERS | SHARE OF SPEND | DAYS SINCE PURCHASE | ORDERS | HISTORIC SPEND | AVERAGE ORDER | DAYS AS CUSTOMER |
|---|
A real retailer, with the edges left visible
Online Retail II · Daqing Chen, UCI Machine Learning Repository.
1,067,371 original line items from a UK non-store retailer. Many customers are wholesalers. Original coverage: 1 December 2009 to 9 December 2011. Meridian is a fictional brand used for this demonstration.
Open the original dataset ↗Computed from the same ledger as the charts
Inspect the SQL behind the dashboard