Checkout conversion: the ten experiments behind +11%
The team was optimising the wrong step. Payment was instrumented to the pixel, and payment was fine. What nobody had measured was the gap: 38% of the people who eventually converted had left checkout at least once and come back, and when they came back the product treated them like strangers. No cart memory, no "you were here", a fresh start at step one.
That number came out of session recordings and a funnel rebuilt by user rather than by session. It changed the question from "how do we make payment faster" to "how do we make returning cheap". Ten experiments later, across four EU markets, checkout conversion was up 11%.
How the programme was run
Every experiment had a one page hypothesis before anything was built: what we believed, what would change, the metric, the sample size, the date we would read it. No peeking. No calling winners at 80% confidence because the quarter was ending. The losers were as useful as the winners, because each one closed a door the team had been arguing about for a year.
What moved the number
- An honest timer. “We’ll hold your spot for 30:00”: the real hold time, framed as a service, not a threat. The single largest lift, +3.1% payment completion (EXP-01).
- Unbroken context. Three page loads collapsed into two named steps, with timer, progress and order summary persisting across them. Step drop-off down 18% (EXP-02).
- Price without surprises. An itemized total with “all taxes and fees included”, and free cancellation restated in the summary, where the decision is made (EXP-03, EXP-04).
- Fewer fields, asked later. SSO first, email first, account creation deferred until after payment (EXP-05, EXP-09).
What did not
- A price expiry countdown, “the price rises when the timer ends”. It won the click and lost the trust guardrail; killed (EXP-06).
- A fully single page checkout. Cleaner on paper, 1.2s slower on mid range Android, and mobile conversion fell 0.9%; killed (EXP-08). Its principle came back as the two named steps.
- Email first, on its own. Flat until it was paired with SSO and a post purchase account prompt (EXP-09).
- Payment methods ordered per market. A clear win in 2 of 4 markets and flat in the rest, so it rolled out per market, not globally (EXP-10).
The lesson I reuse on every funnel
Instrument the gaps, not the stages. Most analytics setups count people at each step and lose them between steps: the back button, the tab switch, the return visit three hours later. The leak is almost always there, and it is invisible until you rebuild the funnel per user. I have found the same shape since in a co-living booking flow (+34% booking conversion) and in a B2B onboarding. Different products, same blind spot.
The full case with all ten experiments is here: GetYourGuide checkout redesign. If your funnel has a leak you cannot find: how a CRO engagement runs.
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.