Instrument the gaps, not the stages: funnel analytics that find the real leak
Every analytics setup I have inherited counts people at each step of the funnel. Product page, cart, checkout step one, payment, confirmation. The numbers are right and the picture is wrong, because the leak is almost never inside a stage. It is between stages, in the moments the stage counters cannot see.
What stage counters miss
- The back button. A user reaches checkout, goes back to compare, returns. Two checkout entries, one person, and a stage funnel counts a drop that did not happen and misses the friction that did.
- The tab switch. Opens the price in a new tab, opens the competitor, comes back forty minutes later. Session timed out; a new session starts at step one.
- The return visit. Comes back the next day with no memory of the cart. At GetYourGuide this was the number that changed everything: 38% of the people who eventually converted had left checkout and come back, and the product treated them like strangers.
Rebuild the funnel per user
The fix is a query, not a tool. Take every user who reached the final step in a period, and walk their events backwards: how many checkout entries, how many hours between the first and the last, which step they left from each time. Then do the same for users who did not convert. The gap between the two populations is your leak, and it is usually not a step. It is a re-entry with no memory, a price that changed between visits, a form that cleared.
Three dimensions make the query useful: entries per converter (if the median is above one, re-entry is your biggest surface), time to convert (if it is hours, the session model is lying to you), and exit step on the first visit (which tells you where to put the memory).
Then instrument what you found
Only now add events, and add them for the gaps: a checkout_reentry event with the step and the hours since last exit, a cart_restored event, a price_changed_since_last_visit event. These are the events that will tell you whether the fix worked, and they are the ones no default setup ships with.
What this looked like twice
At GetYourGuide it became ten experiments and +11% checkout conversion. The winners all closed the gap: an honest hold timer, two named steps with the order summary persisting between them, and the math shown at the total. At a co-living platform the same query found the drop between application and booking was people returning after the roommate conversation, and the work that followed took booking conversion up 34%. Different products, same blind spot. The ten experiments are written up separately.
A note on tools
Any analytics product can do this if it exports raw events; most cannot do it in their funnel UI, which is why the funnel UI keeps telling you payment is the problem. Session recordings are the fastest way to see a re-entry with your own eyes; the per user query is how you know how many there are.
If your funnel has a leak you cannot find: how a CRO engagement runs. The case: GetYourGuide checkout redesign.
Copy the checklist: Checkout audit checklist, with a prompt to use it and its changelog.
Working on a funnel like this?
Finding the leak and proving the fix is what the conversion work looks like. Explore a conversion opportunity → Or email me.