A single run answers “is this push safe?” The Trends tab answers “is quality drifting?” — folding every run in a project into time buckets so you can see verdicts, regressions, and cost/latency move across pushes.

Open it from the Trends tab on any project (next to Runs, Pull Requests, and Settings).

## The charts

• +Regression rate. Area chart of the mean regression rate per bucket, plus total blocking regressions — is quality drifting over time?
• +Verdict history. Stacked area of pass / conditional / fail / inconclusive counts per bucket — the shape of your migration confidence.
• +Cost / latency. Line chart of mean cost and p50/p95 latency deltas per bucket — are newer runs getting cheaper or slower?
• +Slice heatmap. A grid of slices (rows) by bucket (columns); each cell is colored from its mean pass rate — green toward 100%, red toward 0%.

## Slice explorer

Below the heatmap, the slice explorer drills into one slice: pick a slice from the Slice selector (or All slices) and you get its heatmap row plus a pass-rate-by-bucket table. Useful when one cohort — say a hard-cases slice — is regressing while the overall number looks fine. Slices come from your config; the implicit all slice always exists.

## Filters live in the URL

Branch, bucket size (day / week), and the selected slice are all stored as URL query params, so any view you reach is a shareable, bookmarkable link. Defaults: all branches, daily buckets, no slice selected.

## Notes

• +Trends read only denormalized run columns and the per-run summary — never the run blob — so the page stays fast as history grows.
• +Per-slice deltas fold by slice name. Renaming a slice between runs shows up as one slice dropping and another appearing, not a move.