funnel

Also called: funnels, conversion funnel, funnel analysis, funnel conversion rate

A sequence of steps users should go through, with the share who reach each step. Funnel results depend on step order, the conversion window and who enters.

A funnel follows users through an ordered list of steps that should lead to a goal: sign-up → create a group → invite someone → add a shared expense, or product page → cart → payment. For each step it shows how many users reached it, so you can see where people drop off and which step to fix first.

Two numbers describe every step:

$\text{step conversion}_k = \frac{n_k}{n_{k-1}}, \qquad \text{overall conversion}_k = \frac{n_k}{n_1}$

where $n_k$ is the number of users who reached step $k$ . Step conversion points to the weakest link; overall conversion tells you what share of everyone who entered got that far. Look at absolute losses too: a step with a mediocre rate but many users entering it can cost more people than a step with a terrible rate further down.

A funnel isn't just a list of counts; three settings change the answer:

Step order. A strict funnel counts a user at step 3 only if they did steps 1 and 2 first, in that order. A loose one counts anyone who did step 3. Strict is the default for onboarding.
Conversion window. How long a user has to get from the first step to the last (for example 7 days). A short window undercounts slow users; an open-ended one makes recent users look worse, because they haven't had time yet.
Who enters. Which users and which dates form step 1: all sign-ups, only organic ones, only users on the latest app version. Mixing traffic sources in one funnel can hide a problem in one of them.

Use funnels for flows with a clear goal and a natural order: onboarding, checkout, upgrading to a paid plan. For behavior that repeats over weeks, such as whether people come back, use retention and cohort tables instead.

Example

Halves looks at the onboarding funnel for 2,000 users who signed up in one week, strict order, 7-day window:

Step	Users	Step conversion	Overall conversion
1. Signed up	2,000	—	100%
2. Created a group	1,300	1,300 / 2,000 = 65.0%	65.0%
3. Invited a member	910	910 / 1,300 = 70.0%	45.5%
4. Invite accepted	546	546 / 910 = 60.0%	27.3%
5. Shared expense added	455	455 / 546 = 83.3%	455 / 2,000 = 22.75%

The lowest step conversion is 3 → 4 (60%): invites that nobody accepts. But the biggest absolute loss is 1 → 2: 2,000 − 1,300 = 700 users never create a group, versus 910 − 546 = 364 lost at the invite step. Raising step 2 from 65% to 70% would add 100 group creators; if the later rates hold, that's 100 × 0.70 × 0.60 × 0.833 ≈ 35 more activated users a week. Maya now knows where both candidate fixes sit and roughly what each is worth.

Common mistakes

Fixing the lowest percentage by reflex. Compare absolute losses and how many users each fix would really add.
Counting recent users without a full window. Users who signed up yesterday haven't had 7 days yet; leave them out or the last step looks broken.
Loose order where order matters. Users who added an expense before creating a group (for example, invited into someone else's group) inflate a strict onboarding funnel if you count them anyway; decide how to treat them.
One funnel for all traffic. Invited users and paid-social users behave differently; split by acquisition channel before drawing conclusions.
Treating the funnel as the cause. A funnel shows where users leave, not why; pair it with session recordings, interviews or an experiment.

Learn it in the course

Funnels: conversion window, step order and who counts · How a funnel conversion rate is computed, and why the conversion window, the step order and who enters give different numbers on the same users.
Case: fix the onboarding funnel · Find where new Halves users drop off and what to fix first.
Reading a funnel: right censoring and date ranges · Right censoring (users who haven't had time to convert yet), date ranges and other ways a funnel conversion rate misleads.