Funnels locate where people leave, not why

What it is

A conversion funnel writes a process as ordered steps and marks where counts collapse. It answers which step loses people. It does not answer why they leave. A payment-step crash can be price, trust, a technical failure, a missing method, or comparison shopping; the funnel compresses those possibilities into one gap. Treating the gap as a cause confuses localization with attribution.

Why it happens

A funnel is conditional counting: of those who reached a step, how many fired the next event. The difference is a residual, not a cause. Mutually exclusive mechanisms can produce the same residual—an interface barrier, a finished task, an interruption, a fraud check, or a definition that credits completion elsewhere. Funnels also impose a linear order on paths that backtrack, skip, and fork. The largest gap therefore says that step’s pass condition is the harshest or the easiest to bypass, not whether copy, performance, price, or the flow itself should change.

Studying it

Treat the funnel as a locator, not an explanatory model. For the largest gap, list rival accounts in advance and attach an observation that could discriminate among them: error codes and timeouts, field-level abandonment, back-navigation, support themes, concurrent price or inventory shifts. Session replay or task tests separate “saw it and did not act” from “never saw it.” When several accounts predict the same leak location, stop adding steps and move to experiments, interviews, or event splits.

Where it stops holding

Funnels help when the process is strongly ordered and completion is well defined—checkout, sign-up, an application. Exploratory browsing, multi-goal sessions, and content that users may legitimately leave will mark ordinary exits as leaks. Bots, retries, and cross-device continuation distort step counts. Where the leak sits also moves with traffic mix: a step crowded with first-timers is not therefore hard for returning users. A funnel also cannot estimate what was lost among people who never entered the first step.

Applying it

• Report arrivals, conversion, and drop-off at each step, and treat the largest gap as a starting point for inquiry.
• Write mutually exclusive causes for that gap and specify which logs, replays, or experiments would rule each one out.
• Do not ship a color or tooltip change from “this step converts poorly” alone; first show that people reached the step and attempted it.
• On retest, ask whether the hypothesized mechanism declined, not only whether overall conversion recovered.

Related

• Same group: Q3.12.2 Retention curve shape outweighs a single-day number · Q3.12.3 Cohort analysis keeps new and returning users apart · Q3.12.4 Inconsistent step definitions make conversion incomparable · Q3.12.5 Merging entry paths hides a path’s true conversion · Q3.12.6 Survivor bias overstates typical long-term experience · Q3.12.7 Rolling and classic retention are not interchangeable
• Adjacent: Q3.11 Event logging and instrumentation · Q3.08 Error rates and help-seeking
• Search terms: conversion funnel · drop-off · attribution vs localization