methodData, experimentation & growth

Funnel Analysis

Funnel analysis measures how eligible users progress through a defined sequence of events and where they stop or diverge.

Download Markdown (.md)

01What it means

Funnel analysis represents a journey as ordered, instrumented steps and calculates how many eligible users or accounts reach each one. A closed funnel requires entry at the first step; an open funnel may include people entering later. Funnels can enforce order, time limits, and exclusion rules, so the exact configuration is part of the metric rather than a display preference.[1][2][3]

02Why it matters

A funnel turns an overall conversion result into stage-level evidence. It can identify where progress changes by device, cohort, or use case and help teams focus research on consequential friction. Used carefully, it also provides a shared operational definition for experiments involving signup, onboarding, checkout, publishing, or renewal.

03When to use it

Use funnel analysis when the intended outcome has a reasonably defined sequence and events can be measured reliably. It works best for bounded tasks, not exploratory journeys with many legitimate paths. Pair quantitative drop-off with usability research or session evidence before deciding that every exit is a defect.

04How it works

  1. Define the eligible population, identity unit, ordered steps, completion window, entry rule, and whether steps may repeat or occur between named events.
  2. Audit event semantics, timestamps, deduplication, cross-device identity, and missing data before interpreting conversion.
  3. Calculate unique units reaching each step and the conversion and abandonment between steps.
  4. Break down results by relevant cohort or property while maintaining adequate samples and guarding against multiple comparisons.
  5. Investigate high-impact gaps, test a specific change, and monitor downstream quality and guardrail metrics.

05A practical example

Example

A marketplace follows 1,000 eligible mobile users from listing creation: 620 upload a photo, 550 reach preview, and 500 publish. The visualization shows that photo upload loses the most users, while the full journey converts 50%. Research reveals unclear permission recovery rather than unwillingness to list. A redesigned recovery path improves publication, while duplicate listings and moderation failures remain stable as guardrails.

Journey snapshot · illustrative data Where do users leave the listing flow?
  1. 01Start listing1,000
    100% of startersEntry point
  2. 02Upload photo620
    62% of starters62% from prior step
  3. 03Reach preview550
    55% of starters88.7% from prior step
  4. 04Publish listing500
    50% of starters90.9% from prior step
380 users lost at photo upload50% end-to-end conversion

Counts reveal the biggest loss; rates make stages comparable. Neither explains why people leave.

06Where it breaks

Funnels simplify real behavior and can hide loops, alternate routes, offline steps, and reasons for exit. Event loss or identity stitching may manufacture drop-off. A high conversion rate can still produce a poor or coercive experience, and optimizing one step may reduce downstream quality. Observational funnel differences do not establish why users behaved differently.

Common mistakes

  • Choosing steps because they are already tracked rather than because they represent the user journey.
  • Comparing funnels with different windows, entry rules, or identity units.
  • Assuming the largest percentage drop is the highest-value or most fixable opportunity.

07Sources

  1. Funnel explorationGoogle Analytics Help · Authoritative analytics documentationFunnel exploration
  2. Google Analytics Data API funnel reportsGoogle for Developers · Official API documentationGoogle Analytics Data API funnel reports
  3. Build a funnel analysisAmplitude · Product analytics documentationBuild a funnel analysis
View saved topics

Start typing to search titles, definitions, categories, and frameworks.