← All research

Business metric → product outcomes

Metric: % of users who purchase ≥1 wishlisted item within 30 days of saving. Decomposed into a six-stage behaviour funnel. The engine's evidence places the largest addressable loss at Stage 3–4 (Evaluate & Resolve): friction concentrates there (57.7%), and its causes are personal per user — so the product must ask and route rather than treat every save identically.

1 · Save

Observed:
User hearts a product.
Breaks when:
Save ≠ intent: comparing, occasion, fit doubt, social check, inspiration
MVP lever:
Reason-for-save prompt (optional, skippable) separates genuine intent (80%) from noise

2 · Return

Observed:
Reopens wishlist or not.
Breaks when:
Static list gives no reason to come back; intent expires quietly
MVP lever:
Non-monetary change alerts: size back, new reviews, delivery improved

3 · Evaluate

Observed:
Opens a saved item.
Breaks when:
57.7% of corpus friction sits here — the item can't answer 'why buy now?'
MVP lever:
'What changed' cards + routed uncertainty-resolver modules

4 · Resolve uncertainty

Observed:
Checks size, reviews, delivery, returns.
Breaks when:
Fit is the top info gap (21.1%); trust is the top friction overall (reach 25.5%)
MVP lever:
Fit-confidence sheet + trust & return-certainty sheet

5 · Move to Bag

Observed:
Adds saved variant to bag.
Breaks when:
Variant lost between save and bag; OOS kills momentum
MVP lever:
Variant memory + recovery paths (swap size / alert / similar)

6 · Purchase

Observed:
Completes order.
Breaks when:
Checkout friction & final confidence dips
MVP lever:
Out of MVP scope; monitored via guardrails

Why routing beats one global fix: the engine ranks Trust (#1, opportunity 0.256) above Returns (#2, sev 4.67) above Fit (#3), but segment lift shows different users are blocked by different frictions (new/low-trust 2.81× trust; fit-uncertain 3.99× fit; bulk-returners 4.99× returns). A single fixed feature would help only one slice — the router lets each save declare (or reveal) which blocker applies.