Conversion Funnel

The conversion funnel exists only as a mental model in the growth lead's head. When someone asks 'what's our signup-to-paid conversion rate?' the answer is a guess — 'somewhere around 3%, I think.' There's no written definition of funnel stages, no agreed-upon stage boundaries, and no documentation of what counts as a conversion event versus a drop-off.

None — AI cannot analyze conversion patterns because no conversion funnel definition or stage documentation exists anywhere in the organization.

Conversion funnel stages are documented in scattered places — a slide deck from last quarter's board meeting, a product spec that defines onboarding steps, and a marketing wiki page listing acquisition stages. Each uses different stage names and boundaries. The growth team's funnel has seven stages; the marketing team's has four. Nobody agrees on whether 'activated' means 'completed onboarding' or 'used a core feature.'

AI could potentially extract funnel stage names from existing documents, but cannot build a coherent conversion model because the scattered definitions contradict each other on stage boundaries and conversion criteria.

The conversion funnel is documented in a shared analytics wiki with defined stages: visitor, signup, activated, engaged, trial-to-paid, and expanded. Each stage has entry criteria and the tracking events that mark transitions. The growth team references this document when building dashboards. But the funnel definition doesn't link to the actual event taxonomy in the analytics platform, so discrepancies between the documented funnel and measured funnel go undetected for weeks.

AI can parse the documented funnel definition and generate stage-by-stage conversion reports, but cannot validate that measured conversion rates align with the intended funnel design because the documentation isn't linked to the underlying event instrumentation.

The conversion funnel is a well-maintained, current document that maps each stage to specific tracking events, user properties, and measurement queries. A product manager can ask 'what's the drop-off rate between activation and first value moment for enterprise accounts this month?' and get an accurate answer because the funnel definition, segment definitions, and measurement logic are all aligned and findable. Stage definitions update when the product changes.

AI can generate conversion analysis by segment, identify the highest-friction funnel stages, and flag anomalous drop-off patterns. Cannot yet model alternative funnel paths because the documentation captures only the primary linear funnel, not branching user journeys.

The conversion funnel is a formal entity in a product analytics ontology. Each funnel stage has validated relationships to tracking events, user segments, feature areas, and experiment variants. Multiple funnel definitions coexist — primary acquisition funnel, expansion funnel, reactivation funnel — each with machine-readable stage graphs. An AI agent can query 'which funnel stages have the highest variance across enterprise versus SMB cohorts when filtered by signup source?' and traverse the ontology to answer.

AI can autonomously construct custom funnel analyses, compare conversion paths across segments, detect funnel degradation in real-time, and recommend experiment targets for the highest-impact drop-off points.

Conversion funnel definitions are self-documenting and self-updating. When the product team ships a new onboarding flow, the funnel schema automatically detects the changed user path, proposes updated stage boundaries, and validates them against incoming behavioral telemetry. New funnel variants for emerging user segments are generated automatically. The funnel model is a living representation of how users actually progress through the product.

Can autonomously maintain, evolve, and optimize conversion funnel definitions. AI detects shifts in user behavior, proposes funnel redesigns, validates them against real conversion patterns, and publishes updated funnel schemas — all without human documentation effort.

Ceiling of the CMC framework for this dimension.