3 Apr 2026 · Journal
Cohort replay versus the funnel chart
A funnel chart is a compression. Compression is useful until it erases the loops, backtracks, and skipped steps that actually explain the number.
In Journey Signal Lab we still draw funnels — once — so the room can see what leadership already believes. Then we replay the same cohort in order. The replay almost always shows people entering “step three” before “step two,” or repeating a verification screen that the funnel counted only once.
Order matters because stage contracts depend on it. If your atlas says onboarding starts after account creation, but half the cohort created an account from a deep link mid-task, your first stage is a fiction. Replay makes that visible without a data-science project. You are simply watching sequences instead of buckets.
Loops are not noise
Product teams often ask us to “exclude outliers who went back.” Those people are frequently the ones hitting a broken save state. Path Density Workshop exists to decide whether a loop is rare-and-expensive or common-and-harmless. A funnel cannot tell you; it has already flattened the loop into a single pass.
Eligibility still applies. Replay of “everyone who visited pricing” will include staff, bots, and people who cannot buy in their country. We filter the cohort before we replay. Otherwise you get a cinematic mess that feels insightful and proves nothing.
None of this means funnels are banned. They are a headline. User Journey Mapping Analytics needs the article underneath: ordered events, named stages, and a denominator you can defend in critique. If your tool only offers funnels, export a sample and replay it in a spreadsheet. The humility is the point.