InsightPilot

Why the return rate moved — rate effect vs mix effect, rebuilt weekly from BigQuery.

Explain by

Headline

What moved the rate

Pushes the return rate up Pulls the return rate down Net change

Week by week, this year against last

Four-week trailing average, so a single noisy week doesn't read as a trend. Both lines cover 52 weeks aligned to the same position in the year.

This year Last year

Segment detail

How the split is calculated

The overall return rate is a weighted average: each segment's own rate, weighted by its share of items. So a change in the total can only come from two places — the shares moved, or the rates moved. The split is exact and leaves no residual:

Δ rate = Σ (share_now − share_before) × rate_before ← mix + Σ share_now × (rate_now − rate_before) ← rate

Mix is what you get if every segment had held its old return rate and only the traffic composition changed. Rate is what you get if the composition had held still and only behaviour changed.

The split depends on what you segment by. A shift that is invisible inside one dimension shows up clearly in another, which is why the control at the top changes the answer and not just the chart.