Funnel Analysis Basics
Measuring the road users actually take
A funnel is a sequence of events users pass through toward a goal. Funnel analysis measures conversion at each consecutive step to locate where users abandon.
What you will learn
- Define a funnel as an ordered list of events with real conversions
- Compute step-by-step conversion rates
- Explain what funnels reveal and what they hide
A funnel is signup → activation → first value → paid, with conversion measured between consecutive steps. The step with the biggest drop-off is where you fix things first.
What is this?
A funnel is an ordered sequence of user actions leading toward an outcome — for example visit checkout → enter shipping → enter payment → place order. Funnel analysis computes, for each step, how many users who reached the previous step also performed this one, exposing the conversion rate and the drop-off per step. Its value is diagnostic: the largest single drop-off between adjacent steps points to the friction that costs the most conversions.
Why do you need it?
- Overall conversion hides where users quit; step-level conversion shows the exact bottleneck.
- Funnels make UX intervention measurable: fix a step, the funnel tells you if it worked.
- Comparing funnels across channels/cohorts shows where segments differ.
- Number declension: 'checkout converts at 41%' is actionable where 'we lose revenue' is not.
Mental model
A stadium's entrance gates
A funnel's steps
A stadium that holds 50,000 but has one slow turnstile blocks a stadium's worth of fans. Funnels find the turnstile — the gate with the biggest gap between people heading in and people getting through.
How it works
- Pick the outcome and reverse-engineer the prerequisite steps (checkout needs cart, cart needs search/listing, etc.).
- Define each step as an exact event (checkout_started, payment_submitted, order_confirmed).
- Count unique users who reached each step, in order; a user who skips a step counts at the next eligible step (ordered vs open funnel).
- Compute step conversion = users reaching step k / users reaching step k−1 (or / step 1 for overall).
- Identify the step with the largest relative drop and dig into its details.
A simplified checkout funnel.
In production
- An e-commerce team finds the payment-step drop is twice any other and discovers a geoblock template killing international cards.
- A SaaS monitors its signup→activation funnel weekly; any step-level regression pages the owning team within a day.
- A market analysis compares onboarding funnels by device and rebuilds the Android flow around the leaky step.
Best practices
- Define steps with explicit events and an explicit order rule (open vs closed funnel) and write it down.
- Compute deduplicated user counts per step, not raw events — double-fired events fake conversions.
- Use step-conversion AND overall-conversion reporting; the two answer different questions.
- Set funnel baselines before redesigns so the change is measurable.
Common mistakes
- Requiring strict order when users legitimately skip steps (counting them as failures).
- Measuring funnel steps as page views on a client-rendered app, where the 'page' fires for bots and pre-render.
- Putting 12 steps in the funnel so every step looks like a leak, buzzing the team constantly.
- Comparing funnel numbers across populations (new vs returning) without saying which is which.
Debugging tips
- A step's conversion > 100% means the previous step was undercounted (different event sources) — reconcile both steps first.
- Abrupt step drops often follow tracking deploys; check the step's event volume before theorizing about UX.
- For gaps between page-view and action events, use the action event as truth, not the page.
- Segment the funnel before investigating: a leak on one platform hides in the aggregate shape.
Interview checkpoint
Quick quiz
The step with the largest relative drop-off is usually:
Raw event counts are a fine substitute for unique-user counts in funnels.