Funnel

Origin & evolution

The conversion-funnel idea descends from the AIDA model of advertising attention (Awareness, Interest, Desire, Action), credited to Elias St. Elmo Lewis around 1898 and refined through the twentieth century. The version that product and growth teams use today was sharpened by Dave McClure in a 2007 Seattle Ignite talk, Startup Metrics for Pirates. He compressed the customer lifecycle into five stages, Acquisition, Activation, Retention, Referral, Revenue, and the initials spell AARRR, hence "pirate metrics🔢MetricStrategyA unified metric that measures progress, health, or behaviour across the productView reference →".

McClure's point was disciplinary. He wanted founders measuring whether they activated, retained, and monetised users, not just counting eyeballs at the top. The framework became page one of the lean-startup playbook📒PlaybookCustomer SuccessA standard operating procedure for CSView reference → because it turned a vague sense of "we need🫀NeedUserA user need, pain, desire, or constraintView reference → more users" into a sequence of conversion rates a team could attack one at a time.

The honest critique came later. Reforge argued that the fastest-growing products are better described as loops than funnels: a funnel pours users in at the top and out at the bottom with no mechanism to reinvest the output, so it never compounds. The field's settled position treats the two as complementary. A funnel is the right lens for diagnosing where a single pass loses people; a growth loop♾️Growth LoopGrowthA self-reinforcing growth cycleView reference → is the right lens for understanding how the product feeds its own next pass.

How it works in practice

A B2B onboarding tool defines a five-step activation funnel: lands on signup (10,000), creates an account (4,200), connects a data source🛢️Data SourceData & AnalyticsA data source or integrationView reference → (1,900), invites a teammate (980), reaches the "aha" dashboard🎛️DashboardData & AnalyticsAn analytics dashboardView reference → (610). End-to-end conversion is 6.1 percent. The steepest single drop is signup to account, where 58 percent vanish.

The team resists the urge to widen the top with more ad spend. The funnel says acquiring more landers wastes money while 58 percent bounce at the first form. They cut the signup form from nine fields to three and defer the rest into the product. Account creation climbs from 4,200 to 6,100, and because the later steps hold their rates, the "aha" cohort👥CohortGrowthA group of users sharing a common characteristicView reference → grows from 610 to roughly 880 on the same traffic. The funnel earned its keep by pointing the fix at the step that mattered.

Funnel vs. its neighbours

• Growth loop is a closed cycle where an output, a new user or a piece of content, becomes an input that drives the next cycle. A funnel is a one-directional path with no feedback. You diagnose a single conversion path with a funnel; you model compounding growth with a loop.
• Cohort groups users by when they arrived and tracks them forward over time. A funnel measures conversion across stages at a moment. A funnel asks "where do people drop?"; a cohort asks "do this month's users behave like last month's?".
• Customer journey maps the lived experience across touchpoints👆TouchpointCustomer SuccessA customer interaction touchpointView reference →, including emotion and intent, often non-linear. A funnel abstracts that experience into countable stages. The journey is qualitative and wide; the funnel is quantitative and narrow.

In the graph

In the Unified Product Graph, a funnel sits in the Growth domain within the Business, GTM and Growth region, where it is a root. It owns its stages through funnel_contains_funnel_stepFunnelcontainsFunnel Stephierarchy and binds to the people it measures via funnel_maps_personaFunnelmapsPersonacross-domain. Upstream, a product_measures_funnelProductmeasuresFunnelhierarchy edge ties it to the product, and marketing_channel_drives_funnelMarketing ChanneldrivesFunnelcross-domain feeds it from the demand side. The structure enforces the canonical anti-pattern guard: a funnel with no steps is incomplete by construction, because the steps are where the drop-off, and therefore the insight🔦InsightUser ResearchA synthesised finding from researchView reference →, lives.