Mixshift.

Both groups can prefer A, while the overall result prefers B. Move the traffic sliders and watch the average change its mind.

01 · Choose a setting

02 · Change the mix

Who gets the harder cases?

Each slider sets the share of easy cases within one option. Rates within each cohort stay fixed.

Option A · easy-case share

The remainder are hard cases.

Option B · easy-case share

The remainder are hard cases.

03 · Compare the evidence

Same rates. Different average.

Option AOption B

Synthetic teaching data, not a measurement or recommendation. Overall rates are weighted averages: easy share × easy rate + hard share × hard rate. Each option has its own adjustable cohort mix; this compares weighted rates rather than raw counts.