Metrics & Kill-Criteria

The point of this doc is to make the plan falsifiable. A plan you cannot fail is a wish.

North-star

Does the retention curve flatten? Across every platform we studied, the defining signal of product-market fit was a cohort whose return rate stops declining and holds. A curve that decays to zero means no fit, regardless of how big the top of the funnel is. This is the one number that matters most.

Two funnels, measured separately

Player funnel (the majority):

• Activation: reached the first "something fun to do" moment in session one.
• Retention benchmarks (industry rough guide): D1 ~40%, D7 ~20%, D30 ~10%. D1 50%+ is the modern bar. D30 is the predictor of long-term health.
• Session length target: 15+ minutes average.

Creator funnel (the thin layer):

• Activation: first object placed in under 60 seconds.
• Creator retention: returns to edit/build in week 2 and week 4. Target creator D7 > 25% before we build any marketplace/matching features. Supply-side retention is the gating metric; ~67% of failed marketplaces die on supply.
• Conversion: % of visitors who place at least one object (this is a creator-funnel metric, NOT the top line).

Vanity metrics we will explicitly ignore

• Registered accounts / signups. Decentraland had tens of thousands of "monthly" accounts and ~30-38 on-chain DAU.
• Valuation or press impressions. Decentraland and The Sandbox were each ~$1.3B with hundreds of real daily users.
• Cumulative "worlds created." Horizon Worlds had >10k worlds; <9% were ever visited by 50+ people.
• Total tips/GMV at small scale. Rounds to zero at alpha scale; not a retention signal.

Synthetic-supply discipline

• Cap AI-seeded content at ~30% of visible/active content.
• Convert seeded content to real creator content within ~60 days.
• Track the seed-to-real conversion rate explicitly. If seeded content is consistently ignored (low visit/interaction vs creator content), that is a signal the seeding is filler, not fun.

Decision gates

Phase 1 to Phase 2 (open beta): GO only if a cohort returns unprompted in week 2 (curve shows early flattening), first-object < 60s for most, 50 concurrent stable, zero crashes in 48h. Otherwise PIVOT (fix loop) or HOLD.

Phase 2 to Phase 3 (GA + marketplace): GO only if player D7 holds and the retention curve flattens at a non-trivial level across beta cohorts, creator D7 > 25%, and discovery-driven visits are climbing. Otherwise do not build the marketplace yet; a marketplace on top of a leaky funnel just adds surface area to an empty world.

KILL / RETHINK signals (any phase):

• Retention curve declines to near-zero across consecutive cohorts (no PMF).
• Players who never build do not return (the play-first thesis is wrong; either the loop isn't fun or we're actually creation-first and should admit it).
• AI-seeded content is uniformly ignored (the seeding bet is wrong; pretty is not playable).
• We find ourselves relying on reminder emails / paid acquisition to move retention (masking the absence of intrinsic value).

Write the number down before you look at it. Do not move the goalposts after the cohort lands.