Segmentation · behavioral

Behavioral segmentation

Paste a small table of people or customers with a few structuredbehavioral dimensions — task velocity, interaction count, tenure, spend — and this finds the natural segments. The number of segments is chosen for you by a cluster-quality score (silhouette), and each segment is described by the dimensions that most distinguish it from the average, so the read is explainable rather than a black box. Everything runs in your browser; nothing is uploaded.

This deliberately uses only voluntary, structured signals — no audio, no video, no affect or biometric inference. It also discovers emergent roles from interaction structure (who is a connector vs. a broker) and classifies new records against labeled exemplars; those modes are exposed via the /api/segmentation service.