The Setup Choices Behind Every Funnel Analysis

What is funnel analysis?

In web and product analytics, funnel analysis is the method of defining an ordered sequence of steps, naming a first step that marks entry and a last step that marks completion, then counting how many distinct users reach each step within a fixed time limit. The method turns a stream of undifferentiated events into a shape, because each step holds a subset of the step before it. Nothing in the raw event data decides that shape: an analyst does, by writing the step list, the entry rule and the time limit down.

What makes a funnel a funnel?

A funnel needs three things that a plain event report does not: an ordered step list, a defined entry point, and a conversion window. Order lets the tool subtract one step from the next, entry fixes the denominator every later percentage divides by, and the window decides when an unfinished user counts as lost instead of still in progress. A count of purchase events has none of those three, which is why it can rise in the same week a funnel gets worse.

Ingredient	What it fixes	What breaks without it
Ordered step list	Which event follows which	Steps cannot be subtracted, so there is no drop
Entry step	The denominator for every later step	Percentages have no shared base
Conversion window	When an unfinished user is counted as lost	Recent users drag the number down forever
Counting unit	Whether a row is a user or a session	Denominators across steps stop being comparable

Write those four rules into the funnel's own name. A funnel called "Checkout, user-scoped, strict order, 24h" survives a handover, and one called "Checkout v2" does not.

How does step sequencing change the numbers?

Sequencing decides whether events that arrive out of order or with other events in between still count as a conversion, and each setting produces a different number from identical raw data. PostHog's funnels documentation lists three modes: sequential, where "Step B must happen after Step A, but any number events can happen between A and B"; strict order, where "Step B must happen directly after Step A without any events in between"; and any order, where "Steps can be completed in any sequence." Mixpanel's advanced funnels documentation draws the same line with two settings: "Specific Order Funnels require the user to complete each of the funnel steps in the order laid out", while an Any Order toggle lets users "complete unanchored funnel steps in any particular sequence and still convert."

Take one week of events: 10,000 users reached step 1. Of those, 4,000 reached step 2 at some later point, and 2,600 of the 4,000 reached it with no other tracked event in between. A further 700 users did step 2 before step 1.

Sequencing rule	Users counted as converting step 1 to step 2	Reported conversion
Strict order	2,600	26.00%
Sequential, gaps allowed	4,000	40.00%
Any order	4,700	47.00%

Same events, same week, a 21 point spread. Google's GA4 funnel exploration help page adds a fourth lever on top of ordering, since an open funnel means "users can enter the funnel in any step" while a closed funnel means "users must enter the funnel in the first step." Choose the rule that matches the product, then stop changing it, because a sequencing change and a real conversion change look identical in a trend chart.

What does the conversion window do?

The conversion window is the time a user has to finish the flow after entering it, and shortening it lowers reported conversion without touching the product. Mixpanel's advanced funnels documentation states that by default "all customers have 7 days to complete a funnel from the timestamp they perform the Step 1 event", and that "the maximum amount of time you can choose for the conversion window is 366 days, or otherwise equivalent (12 months, 52 weeks, etc). For session-based conversion windows, the maximum is 12 sessions."

Carry the same 4,000 sequential converters forward. If 2,900 of them finished within 1 hour and 3,600 within 24 hours, the funnel reports three different results from one dataset:

Conversion window	Converters	Reported conversion
1 hour	2,900	29.00%
24 hours	3,600	36.00%
7 days	4,000	40.00%

Match the window to the decision the user is making: a checkout finished in one sitting does not need 7 days, and a paid plan that a buyer takes to a manager does. The same trade-off governs an attribution window, where a longer window credits more conversions and blurs the cause.

Should a funnel count users or sessions?

User-scoped funnels count each person once no matter how many visits the flow spans, and session-scoped funnels count each visit separately, so the two report different denominators for the same traffic. Mixpanel's advanced funnels documentation describes the Uniques method this way: "Users will enter the funnel the first time that they track Step 1 in the time period selected."

Suppose those 10,000 users produced 13,500 sessions that contained step 1, and 3,100 of the 4,000 converters finished inside the session where they started. Session-scoped conversion is 3,100 / 13,500, or 22.96%, against 40.00% user-scoped. Pick session scope when the flow is designed to finish in one sitting, and user scope when it spans visits, then read what a session is before assuming a tool's session boundary matches yours.

One week, six settings, six numbers

Session-scoped22.96%

Strict order26.00%

1-hour window29.00%

24-hour window36.00%

Sequential order40.00%

Any order47.00%

When is funnel analysis the wrong tool?

Funnel analysis is the wrong tool whenever the journey has no fixed order, because the method's entire output depends on an order the users are not following. A dashboard where people use five features in whatever sequence their week demands, and a B2B purchase with four people entering from four channels, fail the same test: no single step list describes how most users move. Forcing a funnel onto those flows charts the analyst's guess about the order instead of the product.

Three signals say to stop. The any-order setting converts far more users than the sequential setting, which means the population is not moving in the claimed order. Most entrants show up at step 3, which means the entry step is not the entry. The flow has no completion event a user could finish, which means there is no last step to divide by. Use customer journey analytics and cohort analysis for those cases.

How does funnel analysis fit next to the other funnel questions?

Funnel analysis is the method, and the neighboring pages cover what you do once the method is set up. Reading the resulting shape and locating where users leave belongs to conversion funnel, the arithmetic of loss per step and end to end belongs to drop-off rate, building and segmenting the report itself belongs to funnel reporting, and the acquisition side belongs to marketing funnel optimization.

What happens after funnel analysis names a bad step?

A funnel names the step that loses users, and a replay of that step names the cause. Flowsery records every user session and detects rage clicks, dead clicks, JavaScript errors, drop-offs and broken flows automatically, then groups matching sessions into one issue and ranks issues by how many users hit them. Its funnel analysis runs on the same event stream as the recordings, so the step and the sessions behind it come from one dataset instead of two tools with two definitions.

Each issue lands in Slack, Linear or Jira with the replay and steps to reproduce attached, and tagging @flowsery in Slack opens a draft pull request in the team's GitHub repo. Flowsery also connects to PostHog or Amplitude replays already recorded, with no re-instrumentation.

Frequently Asked Questions

What is funnel analysis in simple terms?

Funnel analysis counts how many people reach each step of a flow you defined in advance, in the order you defined. The method needs an ordered step list, a first step that marks entry, and a time limit for finishing. Everything a funnel reports comes out of those three decisions plus the events themselves.

Why do two tools report different numbers for the same funnel?

Sequencing rules, conversion windows and counting units differ between tools, and each one changes the denominator. One week of events in the example above reports 26.00% under strict order, 40.00% under sequential order and 47.00% under any order. Compare the two definitions field by field before deciding either tool is wrong.

What conversion window should a funnel use?

Set the window to the longest time a converting user needs for that specific decision, then leave it fixed. Mixpanel's default is 7 days from the step 1 event, with a maximum of 366 days. A shorter window reports lower conversion for the same behavior, so a window change and a product change are indistinguishable in a trend line.

What is the difference between a strict funnel and an open funnel?

Strict order controls what may happen between two steps, while open versus closed controls where a user is allowed to enter. PostHog's strict order mode requires step B directly after step A with no events in between. Google's GA4 open funnel lets users enter at any step, and a closed funnel requires entry at the first step.

How many steps should a funnel have?

Use one step per meaningful state change in the flow, not one per click. Adding steps splits the same loss across more transitions, which lowers each per-step percentage while leaving end-to-end conversion unchanged. That is why end-to-end numbers from funnels with different step counts cannot be compared.

Does funnel analysis work for products with no fixed order?

No. The method reports a sequence, so a journey with no sequence produces a chart of the analyst's assumed order rather than user behavior. When the any-order setting converts far more users than the sequential setting, or most entrants appear at a middle step, switch to journey or cohort analysis instead.

How should a funnel's name describe its setup?

A funnel's name should carry its entry rule, sequencing mode, conversion window and counting unit, because those four choices decide the reported number. A name like "Checkout, user-scoped, strict order, 24h" survives a handover to someone else, while "Checkout v2" hides every setting that produced the result. Write the ingredients into the name instead of leaving them in a config panel nobody checks before comparing numbers across weeks.

What is the difference between a user-scoped and a session-scoped funnel?

A user-scoped funnel counts each person once no matter how many visits the flow spans, while a session-scoped funnel counts each visit as its own attempt. In one week of data, the same conversion measured 40.00% user-scoped and 22.96% session-scoped, because a session-scoped funnel divides by a much larger number of sessions than the count of unique users. Pick session scope for a flow meant to finish in one sitting and user scope for one that spans multiple visits.

When should a funnel analysis be replaced with journey or cohort analysis?

Switch away from a funnel when the any-order setting converts far more users than the sequential setting, since that gap means the population is not following the order the funnel assumes. The same signal shows up when most entrants appear at a middle step instead of the defined first step, or when the flow has no completion event to serve as a last step. Any of those three signals means the chart is showing the analyst's guessed order rather than actual user behavior, and journey or cohort analysis fits better.

What happens after a funnel analysis identifies a problem step?

Flowsery's session replay records every user session and automatically detects rage clicks, dead clicks, JavaScript errors, drop-offs and broken flows, then groups matching sessions into one issue and ranks issues by how many users hit them. Its funnel analysis runs on the same event stream as the recordings, so the step that lost users and the sessions behind it come from one dataset. Each issue lands in Slack, Linear or Jira with the replay and steps to reproduce attached, and tagging @flowsery in Slack opens a draft pull request in the team's GitHub repo.