Isolate novelty effect: ~40% of early UI 'wins' decay to zero within two weeks among returning users.

Isolate novelty effect: ~40% of early UI 'wins' decay to zero within two weeks among returning users.
Returning visitors react to change itself, not to the change being better.
The playbook:
— Segment new vs returning from day one.
— Plot daily uplift over the full run. A spike that trends toward zero = novelty.
— Trust the new-user segment for the durable effect; returning users carry the noise.
— For big redesigns, extend runtime to 3-4 weeks so the curve flattens.
A lift measured in week one is a measurement of surprise, not preference. Read the number, not the story. [decay window ≈14d]