Define a useful return before measuring one.
No conversion lift, retention gain, or research result is claimed. These are proposed experiments with defined cohorts and guardrails, ready for a PM, engineer, and researcher to refine together.
MEASUREMENT PLANUseful discovery
New cooks starting guided cooking within their first two visits ÷ eligible new cooks with two visits or a seven-day window.
- Events to specify
- pantry_started → recipe_viewed → cooking_started
- Experiment
- Three-ingredient setup vs. browsing a curated collection
- Guardrail
- Track no-match rate and ingredient-exclusion failures; irrelevant starts do not count as a win.
MEASUREMENT PLANA repeat cooking habit
Activated cooks completing guided cooking on two distinct days in days 8–21 ÷ activated cooks eligible for that full window.
- Events to specify
- cooking_completed · meal_saved · recipe_returned
- Experiment
- A saved weekly shelf vs. a reminder-only return route
- Guardrail
- Watch notification opt-outs and self-reported ingredient waste; use completion feedback, not screen time.
A small release, with a learning agenda.
01 / VALIDATE THE TASKRun the discovery plan and moderated prototype tasks. Revise the onboarding and recovery paths before building a live service.
02 / PILOT THE WHOLE LOOPRecruit a small, consented pilot. Include the editorial workflow, content operations, and failures, not just the consumer happy path.
03 / COMPARE RESPONSIBLYAgree on cohorts, experiment duration, and a sample-size plan before randomizing. Keep the guardrails alongside the primary metric.
DELIVERED / REFLECTIONA complete product argument, made tangible.
The most useful interface begins before the recipe: it reduces the effort of choosing. The open question is whether a small, relevant set continues to feel helpful as tastes change and pantry data ages.
Delivered: product framing, a research plan, journey and architecture, wireframes, interaction decisions, visual foundations, expert tools, a handoff specification, and a coded mobile prototype.