Maple Market requests descriptive customer groups using fictional visit counts and basket values. Deliver an unsupervised scaling and K-means workflow, indexed assignments, original-unit profiles, group sizes, a silhouette diagnostic and a complete customer-to-cluster handoff. Do not rank customers by arbitrary cluster numbers.
Maple Market wants a descriptive grouping of its fictional customers. The table contains visits and basket values, plus an identifier that must stay outside the distance calculation. There is no outcome label to predict. Build a scaled three-cluster partition, then describe the resulting groups in the original business units.
The marketing analyst needs a row-level assignment table and a group summary, not just a scatterplot or an array of cluster numbers. Cluster IDs are arbitrary, so do not treat their numeric order as customer value. Compute a silhouette score as a geometric diagnostic and count group sizes as a separate check. The final handoff must preserve every customer once. These steps continue the same notebook, adding an output at each stage and keeping the preparation and fitted clustering objects available for inspection.