/funnel-analysis

Funnel Design Guide

Defining funnel steps

A well-defined funnel has steps that are:

1. Sequential — a user cannot reach step N without passing through step N-1
2. Mutually exclusive — each user is counted once per step at a given point in time
3. Exhaustive within scope — no meaningful step is omitted that could explain drop-off
4. Anchored to a denominator — the first step defines the universe; all conversion rates are relative to it

Anti-patterns:

• Steps that can be completed out of order (breaks the funnel assumption)
• Steps defined by page views rather than intentional actions (inflated early steps)
• Mixing session-level and user-level counts across steps

Conversion rate types

Step-over-step conversion rate users at step N / users at step N-1 Answers: "Of the users who reached this step, how many continued?"

Overall conversion rate users at step N / users at step 1 Answers: "Of everyone who entered the funnel, how many reached this step?"

Use step-over-step to find where the biggest drop happens. Use overall to communicate the funnel health to stakeholders.

Time window considerations

Open-ended funnel: A user can complete later steps at any time. Appropriate for purchase funnels where users shop over days.

Time-bounded funnel: A user must complete all steps within a fixed window (e.g., 7 days). Appropriate for onboarding where late completion is not a real conversion.

Cohort-based funnel: Group users by start date and track their progression over a fixed observation window. Required for fair period-over-period comparison.

Drop-off prioritisation

Impact score: absolute drop-off × value per user at that step

A 40% drop-off at step 2 is more impactful than a 40% drop-off at step 5 if step 2 has 10× more users flowing through it.

Recovery value: users lost at step × conversion rate of remaining steps × revenue per conversion

This gives the maximum revenue recoverable if you eliminated the drop-off entirely — use as an upper bound for effort.

Segment-specific funnels

Always break the funnel by the most relevant segments before drawing conclusions:

• Acquisition channel — users from paid ads may have lower intent than organic
• Device type — mobile funnels often drop sharply at form-fill steps
• New vs returning users — returning users bring prior familiarity; comparing them inflates cohort averages
• Plan / price point — high-intent (paid plan) users convert differently from freemium

Funnel quality checklist

• First step denominator is clearly defined and appropriate
• Time window is specified and consistent for all steps
• Users are de-duplicated per step (each user counted once)
• Drop-off at each step has at least one hypothesis
• Biggest drop-off step is identified and owns a next action
• Funnel is broken by at least one key segment