The five funnel analysis mistakes that hide your real drop-off

A funnel answers one question: of the people who started, how many finished, and where did the rest leave? It is hard to build a report that simple and still get a misleading answer, but teams manage it every week. The five mistakes below account for nearly every funnel we have been asked to debug. Each one has a tell you can spot in the chart and a fix that takes minutes.

Mistake 1: counting pageviews instead of people

The oldest funnel mistake is to count how many times a step page loaded rather than how many people reached it. A visitor who refreshes the checkout page three times becomes three checkouts; a visitor who bounces between step two and step three inflates both. The tell is a funnel where a later step is larger than an earlier one, which is impossible when you count people.

Fix: build funnels on visitors (or on identified users, if people log in), not on events. TapCub funnels count each person once per step inside the window, regardless of how many times they trigger it.

Mistake 2: no time window

A funnel needs a window: how long a person has to get from the first step to the last. Without one, someone who saw the pricing page in January and bought in June is a conversion in the same funnel as someone who did both in ten minutes. The conversion rate looks healthy and tells you nothing about the checkout you are trying to fix.

Fix: pick a window that matches the decision. For a checkout, one session or 24 hours. For a free-trial-to-paid funnel, 14 or 30 days. State the window in the report title so nobody compares a 1-day funnel to a 30-day one.

Mistake 3: forcing strict order

Real paths are messy. People open the pricing page, go read the docs, come back, start checkout, check a competitor, return and pay. A funnel configured to require steps in strict sequence with nothing in between will drop everyone who detoured. The tell is a funnel that loses most people at a step where the page itself is fine.

Fix: use “in order, anything in between” as the default. Reserve strict order for flows where the sequence is enforced by the product, like a multi-step form, and for diagnosing whether people skip a step.

Mistake 4: mixing segments and bots

A funnel over all traffic averages together groups that behave nothing alike. New and returning visitors, mobile and desktop, paid and organic, and, most damaging, humans and bots. A crawler that hits every pricing page URL but never starts checkout adds a fat top to the funnel and a terrifying first-step drop. The tell is a first-step drop that is far larger on days with high traffic.

Fix: filter the funnel to humans (TapCub does this by default; the humans-only switch applies to funnels), then break it down by one dimension at a time. The breakdown is usually where the finding is: a 46% drop overall might be 20% on desktop and 70% on mobile.

Mistake 5: reading the total instead of the biggest drop

The overall conversion rate is the least useful number in a funnel. It is a product of every step, so it moves when anything moves and points at nothing. The number to read is the largest single step-to-step drop, because that step is where a change will have the biggest effect on the total.

In the sample funnel below, 2,400 visitors reach pricing, 1,248 start checkout, 674 reach payment and 508 complete. The headline rate is 21%. The useful fact is the 46% loss between checkout and payment, which is where the shipping cost appears.

A five-minute routine

Before you trust a funnel, run through the table below. It takes five minutes and catches all five mistakes. Then make a habit of reading the biggest drop first, breaking it down by device and source, and only then looking at the total. If you want to go deeper on step design, the behavior analytics page shows funnels, paths and retention on the same sample data, and the conversion rate calculator helps you size the effect of fixing one step.

One last habit: save the funnel with its window and filters in the name (“Checkout · humans · 24 h”), and share that saved report rather than a screenshot. Most funnel arguments in meetings are two people looking at two differently configured funnels and assuming they are the same. A named, saved definition ends the argument before it starts, and lets you compare the same funnel month over month without rebuilding it.