← All articles

Analytics · 9 min read · 2026-10-29

Funnel Exit Paths: Where Abandoners Go Tells You Why

A funnel step that loses 40% of users reports one number and hides at least four different problems. The drop-off rate tells you the size of the leak; it says nothing about why users left, because it collapses everyone who abandoned into a single bucket. The signal you are throwing away is the next page each of those users visited. Someone who abandons the trial-setup step and immediately opens the pricing page is telling you something completely different from someone who opens the docs, who is telling you something different again from someone who simply closes the tab. Same leak size, three different causes, three different fixes.

This is the case for treating the abandoner's destination as a labelled friction signal rather than noise. Most funnel work stops at "where do users drop?" and jumps straight to A/B-nudging the leaking step. Exit-path analysis adds one question — "and where do they go next?" — that turns an undifferentiated leak into a ranked set of hypotheses you can act on. The step did not fail for one reason; it failed for several, in proportions the destination breakdown makes visible. This piece is a teardown of the "optimise the leaking step" reflex and a procedure for replacing it with destination-driven diagnosis.

Drop-off is a size; destination is a cause

The core limitation of a funnel chart is that it measures departure, not direction. It tells you how many users failed to advance from step two to step three, and exactly nothing about what those users did instead. Datadog made this explicit when it shipped Journey Paths in 2026, noting that funnels show where users drop but not what they did instead, so "two different experiences can produce the same funnel result" (Datadog). Two cohorts with an identical 40% drop can have wholly different underlying behaviour, and the funnel number flattens that difference into a single percentage.

Destination restores the difference. The next page a user visits in the same session is a behavioural statement about why the step did not work for them. It is not perfectly reliable — a single page visit is a weak signal on its own — but aggregated across hundreds of droppers it produces a stable distribution: X% went to pricing, Y% to docs, Z% to a comparison page, the rest left. That distribution is the thing you actually want to optimise against, because each slice points at a different lever. The drop-off rate tells you a step deserves attention; the destination mix tells you what kind of attention, and in what proportion.

A taxonomy of destinations and what each one means

Four destination classes cover most SaaS funnels, and each maps to a distinct cause. Abandoners who go to the pricing or billing page are signalling cost or packaging friction — sticker shock, an unclear plan boundary, or uncertainty about what the step will commit them to paying. Abandoners who go to docs, a help centre, or an in-app search are signalling a comprehension gap: they wanted to continue but did not understand what the step was asking or how to complete it. These two look identical in the funnel and need opposite responses — one is a pricing problem, the other is a content and onboarding problem.

The other two classes complete the picture. Abandoners who loop back to a comparison, feature, or integrations page are still evaluating — they dropped out of the flow to re-check whether the product fits before committing, which is an unresolved-evaluation signal, not a step defect. Abandoners who leave with no further pageview are the ambiguous group: either genuine low intent or a hard dead end with nowhere obvious to go. That last group is where exit-path analysis hands off to drop-off triage, because "left entirely" still needs the intent-versus-friction split that funnel drop-off triage provides. Exit-path analysis resolves three of the four classes outright; the fourth it isolates so you can triage a smaller, cleaner group.

Why the "optimise the step" reflex misfires

The default response to a leaking step — redesign the step — assumes the step itself is the problem. For destination-labelled abandoners, it usually is not. If most of your droppers bounced to the pricing page, simplifying the form on the leaking step changes nothing about the fact that they balked at cost; you will ship a cleaner form and watch the same users leave for the same reason. If most went to the docs, the fix is inline guidance or a clearer instruction at the step, not fewer fields. The reflex fails because it treats a single step-UX lever as the answer to causes that live outside the step.

This is the same error that makes aggregate funnel numbers dangerous: acting on the blended metric instead of the segments underneath it. A step with a 40% drop that is 60% pricing-driven and 30% comprehension-driven has two real problems and one non-problem (the clean exits), and a step redesign addresses none of them cleanly. Seeing the forward paths is the natural complement to path analysis versus funnels: the funnel tells you the step is leaking, path analysis tells you the shape of what users do instead, and exit-path destination counts turn that shape into a prioritised fix list rather than a pretty Sankey nobody acts on.

Reading destinations without fooling yourself

Before you act on a destination distribution, confirm it is real and not a tracking artefact. The most common failure is an identity break at the drop: if a user moves from anonymous to logged-in, or across a subdomain, and your identity stitching fails, their "next destination" can be misattributed or lost entirely — producing a phantom exit or a bogus path. Verify that sessions are stitched across the abandonment point, that the destination pages actually fire the pageview events you are counting, and that you are measuring same-session next visits rather than conflating a return three days later with an immediate bounce.

Interpret the distribution as hypotheses, not verdicts. A destination tells you what a user did, which strongly implies but does not prove why. The pricing-page jumpers might be comparing plans out of enthusiasm, not balking; the docs-seekers might be power users validating a detail, not confused novices. This is why destination counts pair best with a qualitative layer: pull session replays for the top two destination groups and confirm the story before you build the fix. The quantitative path ranks where to look; replay confirms what you are looking at. Destination analysis narrows four guesses to one or two and tells you which replays are worth watching — it does not excuse you from watching them.

Worked example: the setup step that was really a pricing page

A SaaS team sees a 45% drop between "started workspace setup" and "invited first teammate." The reflex is to streamline the invite step. Exit-path analysis first. Branching forward from the drop-off, the top same-session destinations are: 38% opened the pricing page, 22% opened the "seats and billing" help article, 18% returned to the integrations page, and 22% left with no further pageview. Four groups, four meanings, hiding behind one 45% number.

The fixes follow the labels, in size order. The pricing jumpers (38%) are the biggest group and they are not a UX problem — the invite step implied per-seat billing without saying so, so the fix is a clear "inviting teammates is free on your plan" line at the step, not a cleaner form. The help-article group (22%) is a comprehension gap, fixed with inline copy explaining what the invite does. The integrations loop (18%) are evaluators, handed a short "works with your stack" reassurance. The clean exits (22%) get triaged for intent and traced upstream to the campaign that filled the funnel. Streamlining the invite form — the original plan — would have touched none of the top three causes. Reading the destinations turned one vague "improve setup" into three targeted, non-overlapping fixes ranked by the users they would actually recover.

Make it a repeatable step in your funnel review

Fold exit-path analysis into your standing funnel review as a fixed second question. For every step above a drop-off threshold, do not stop at the leak rate — branch forward, pull the top five same-session destinations, and record each as a percentage of droppers with its implied cause label. The output is a ranked fix list per step: the cause classes sized by how many users each would recover, so you work the largest recoverable problem first instead of the most visible one. This keeps a well-built funnel honest as a diagnostic rather than a scoreboard.

The discipline underneath is simple: a drop-off is a question, and the destination is most of the answer. Teams that stop at the leak rate optimise the wrong step, ship fixes that move nothing, and conclude the step is "just hard" when it was really a pricing signal, a docs gap, and an evaluation loop wearing one number. Spend the extra twenty minutes to read where your abandoners go, and the majority of your funnel "conversion problems" resolve into specific, addressable causes you can rank and ship against this sprint.

Frequently asked questions

The questions below define funnel exit-path analysis, separate it from drop-off and exit rate, explain what different post-abandonment destinations mean, name the tools that show forward paths, and give a procedure you can run on your worst step this sprint.

FAQ

What is funnel exit-path analysis?

Funnel exit-path analysis looks at where users go immediately after they abandon a funnel step, rather than just counting how many left. A standard funnel reports the drop-off rate between two steps — the leak size — but treats everyone who leaked as a single undifferentiated group. Exit-path analysis splits that group by destination: users who jumped to the pricing page, users who went to docs or a help article, users who looped back to a comparison or feature page, and users who left the product entirely. Each destination is a different signal about why the step failed, so the next page a user visits becomes a label for the cause of the abandonment, not just evidence that it happened.

How is exit-path analysis different from drop-off and exit rate?

Drop-off rate and exit rate both measure departure; exit-path analysis measures direction. Drop-off is the share of users who fail to move from one funnel step to the next, and exit rate is the share of pageviews that were the last in a session — both tell you that users left and at which point. Neither tells you where they went. Exit-path analysis adds the next destination, which is the diagnostic piece: a 40% drop where most abandoners open the pricing page implies price sensitivity, while the same 40% drop where most open the docs implies a comprehension gap. The leak size is identical; the fix is opposite.

What do different post-abandonment destinations mean?

Read the top destinations as hypotheses. Abandoners who go to the pricing or billing page suggest sticker shock or an unclear plan — a pricing/packaging problem, not a step-UX one. Abandoners who go to docs, a help centre, or search suggest they wanted to continue but did not understand what the step required — a comprehension or onboarding-content gap. Abandoners who loop back to a comparison, feature, or integrations page suggest unresolved evaluation — they are still deciding whether the product fits. Abandoners who leave the product with no further pageview suggest low intent or a hard dead end. The same drop-off number can be mostly one of these or a mix, and only the destination breakdown tells you which.

What tools show where users go after leaving a funnel?

Product analytics tools with path or journey visualisation, plus session replay for the qualitative layer. Datadog's Journey Paths, released in 2026, lets you open a funnel step, select the drop-off, and compare the sequences converted versus dropped-off users took — including sessions that abandoned before the next step. Mixpanel, Amplitude, PostHog and Heap offer path or flow exploration that can be anchored to a funnel step. Session-replay tools such as Hotjar, FullStory and LogRocket show the struggle behind the path. The key requirement is that the tool can branch forward from a specific drop-off point, not just report the aggregate leak.

How do I run exit-path analysis this sprint?

Pick your worst-leaking funnel step, then branch forward from the drop-off instead of redesigning the step. Pull the top five destinations abandoners visit in the same session after leaving, and compute each as a share of droppers. Label each destination with its implied cause — pricing, docs, comparison/feature, or hard exit — and size the groups. Fix the largest labelled group with its matching lever: pricing clarity for pricing-page jumpers, inline guidance for docs-seekers, proof or comparison content for evaluators, upstream targeting for clean exits. Confirm the destinations are real by checking identity is stitched across the drop so you are not mislabelling a tracking break as a journey.