Open vs Closed Funnels: SaaS Conversion Definitions

Two people can build a funnel on the same SaaS signup cohort and honestly report conversion rates thirty points apart. Neither has a broken pipeline and neither is fudging the number — they have simply made different choices about what a funnel is. Open versus closed, strict order versus any order, and the length of the window between steps each move the headline number, and if you do not know which choices your funnel is using, you do not actually know your conversion rate.

This is the funnel version of the definition trap that already governs retention measurement. It matters just as much, because funnel conversion rates drive decisions about onboarding investment, activation targets, and where to spend engineering time. Get the definition wrong and you either chase a drop-off that is a measurement artifact or celebrate a conversion rate that an over-generous setup inflated. This post lays out the choices plainly, shows how the same cohort produces different numbers under each, and gives a rule for how a SaaS team should define its funnels.

What open and closed funnels actually measure

A closed funnel counts only users who enter at the first step and proceed through the funnel in sequence; a user who lands mid-funnel or completes steps out of order is excluded entirely. An open funnel is more permissive: users can enter at any step and are counted from that point forward, and it typically does not enforce strict ordering. As Analytics Mania puts it, in a closed funnel "a visitor can enter the funnel only in the first step" and cannot join at any later stage (Analytics Mania). That single rule about entry is what separates the two.

The consequence is mechanical and unavoidable: run both definitions over the exact same event data and the closed funnel almost always reports a lower conversion rate, because it throws out users who skipped ahead or backfilled an earlier step later. Open funnels answer "of everyone who touched this step, how many moved on?" while closed funnels answer "of the cohort that started at the top, how many made it all the way through in order?" Both are legitimate, but they are not interchangeable, and quoting one while thinking about the other is how funnel arguments start. This is the same clarity the funnel-tracking playbook insists on before you instrument a single step.

The same cohort, three conversion rates

Take a cohort of 1,000 signups moving through three steps: create workspace, connect a data source, view first report. Suppose 700 create a workspace, 500 connect a source, and 400 reach a first report — but of those 400, some connected their source only after peeking at a sample report, doing the last two steps out of order. A closed, strictly ordered funnel might report 30% (300 users who did all three in sequence within the window). Loosen ordering to any-order and it climbs to 40% (all 400 who eventually did all three). Switch to an open funnel measuring the final step in isolation and the "report viewed" conversion can read 65% or higher, because it is computed against only the users who reached the connect step, not the full cohort.

Three numbers — 30%, 40%, 65% — all describing the same 1,000 signups. This is exactly why cross-company benchmarking is treacherous: a competitor boasting "60% activation" has told you almost nothing until you know whether that is a closed cohort funnel or an open single-step rate, ordered or not, on what window. Compare your strict closed number to their generous open one and you will conclude you are losing when you may be ahead. Before trusting any figure, normalize the definition first — the same discipline the funnel-conversion benchmarks guide applies before it quotes a single percentile.

Step ordering: strict, any-order, and the skip problem

Ordering is a separate lever from open versus closed, and it moves the number just as much. A strictly ordered funnel credits a user only if they completed the steps in the exact sequence you specified; a user who did step three before step two is counted as dropped. An any-order funnel credits anyone who completed all the steps regardless of the path they took to get there. The gap between these two on real data is often large, because users rarely move through software in the tidy line a funnel diagram implies — they explore, backtrack, and complete tasks in whatever order the interface allows.

The right choice depends on whether the sequence is real. When a step genuinely gates the next — you cannot invite a teammate before a workspace exists — strict ordering reflects the product and any out-of-order data is a tracking bug worth investigating. When the steps are independent activation tasks — connect an integration, invite a user, set up a dashboard, in any order — strict ordering invents drop-off that is not really there, and an any-order or checklist funnel tells the truth. Forcing a strict sequence onto a genuinely non-linear flow is one of the quieter funnel-tracking anti-patterns, because the "drop-off" it manufactures sends teams optimizing a step that was never actually the problem.

The conversion window: how long a step has to count

The conversion window is the time a user has to reach the next step and still be counted, and it is part of the funnel definition even though most tools bury it as a setting. A 1-hour window credits only users who advance within the hour; a 7-day window credits anyone who advances within a week. Measuring the elapsed time between steps is itself a useful signal — a long gap flags friction like a verification email or a slow onboarding call — but the window you choose for the funnel determines whether those slow-but-real conversions count at all.

For SaaS this is decisive, because the value action often happens days after signup: a founder connects billing over the weekend, an admin invites the team on Monday, an analyst runs the first real report after a kickoff. A same-session or 24-hour window will silently drop every one of those legitimate conversions and make onboarding look far leakier than it is. Match the window to how long the step realistically takes in your product — for most activation funnels that is measured in days, not minutes — and keep it consistent, because quietly shortening the window between reporting periods redraws the funnel just as a definition change would. This is the funnel analog of the choices covered in attribution lookback and conversion windows, where the window length is the whole ballgame.

Which funnel definition SaaS should use

For the headline activation or trial-conversion number, most SaaS should use a closed, ordered funnel keyed to signup as the entry step, with a conversion window sized to the real time-to-value of the product. This is the honest measure of what share of an acquisition cohort actually reaches first value, and it is the number that belongs in an investor update or a north-star dashboard. Closed-and-ordered is stricter and lower, but it is stable and hard to game, which is precisely what you want from a metric people make decisions and set targets against.

Open funnels earn their keep in diagnosis. When you already know activation is weak and you want to find the specific step that is bleeding, an open, single-step conversion view — of everyone who reached the integration screen, how many finished? — isolates that step without upstream drop-off muddying the read. The practical setup is a two-layer one: a closed cohort funnel for the headline rate you report and benchmark, and open per-step views for debugging where the friction lives. That mirrors the throughline the activation funnel guide builds from signup to first value, where the cohort number and the step diagnostics serve two different jobs.

How the definition quietly breaks benchmarks and dashboards

Because four independent levers — open vs closed, ordering, window, and which events bound the funnel — each move the number, an unstated definition means different people in the company are quietly reporting different metrics under the same name. Marketing pulls an open funnel and celebrates 60%; the product team's closed funnel reads 35%; the board deck splits the difference from a third tool. None of them are lying, but the meeting is now an argument about a number instead of a decision about the product. This is the same failure mode that the metric-definition-drift guide exists to prevent, applied to funnels.

The danger sharpens over time, because funnel definitions drift silently. A tool upgrade flips the default from open to closed, someone shortens the window to "clean up" the data, or a new event is inserted as an extra step — and the conversion rate moves for reasons that have nothing to do with the product. Without a written definition, that movement gets misread as a real regression or improvement, and teams launch investigations or take credit for a change in configuration. A funnel you cannot reproduce from a written spec is a funnel you cannot trust to trend, which undermines every optimization decision downstream of it.

Lock the funnel definition before you optimize

The practical takeaway is to write down the funnel definition and freeze it: open or closed, ordered or any-order, the conversion window between steps, and the exact events that mark entry and each step. Those four choices fully determine the number, and any left ambiguous means your conversion rate is not really a fact about your product — it is a fact about a setting nobody agreed on. A one-line definition in your metrics catalog ends most funnel arguments before they start, exactly as it does for retention.

This discipline pays off twice. It makes external benchmarking honest, because you can restate your number to match whatever definition a benchmark uses instead of comparing apples to oranges. And it makes optimization trustworthy, because a stable funnel is the precondition for reading any experiment: if the funnel's definition can shift underneath a test, you cannot tell a real conversion lift from a configuration change. Teams ready to act on a locked funnel should read the data-driven funnel-conversion guide next, because every lever it pulls assumes the funnel underneath it is defined once and measured the same way every time.