Segmented anomaly detection
Last updated 2026-06-14
Definition
Segmented anomaly detection runs the same checks per slice of your data — by city, channel, or cohort — instead of only on the overall number. A problem in one segment often hides inside a healthy total. Watching each slice on its own means a drop in one channel surfaces even when the headline metric looks fine.
How to do this in Quri
- Pick the segments that matter to you, such as channel, city, or signup cohort.
- Let Quri run detection on each slice in the background, not only the global metric.
- Read the surfaced flags on Pulse to see which segment moved, not only that the total changed.
- Trace a leak in one part of your funnel with the public /tools/funnel-dropoff-finder.
Frequently asked
- Why not only watch the overall metric?
- A healthy total can hide a sinking segment. If one city or channel drops while others rise, the headline number barely moves. Per-segment checks catch the local problem early.
- Which segments can Quri watch?
- Quri detects per slice across the dimensions your sources expose, like channel, geography, or cohort, so a localized issue shows up on its own rather than being averaged away.