Why this matters
Everyone who reaches a checkout has decided to buy, so each person lost there is a real leak. The guide to checkout optimisation lists the usual causes, from forced accounts to late costs, and the guide to product page optimisation shows how doubts left unanswered on the product page surface later as abandoned baskets. This exercise finds the leaks in one real-looking flow and chooses what to fix first.
The checkout walkthrough
Skeinwright Yarns is an invented online yarn shop run by two owners, with a freelance developer for a few days a month. The owners pulled a month of step counts and gathered what they could from recordings, a one-question exit poll on the account step and the support inbox.
Skeinwright's checkout funnel for one month
| Step | Shoppers reaching the step |
|---|---|
| Basket viewed | 7,800 |
| Checkout started | 4,100 |
| Past the account step | 2,700 |
| Past the delivery step | 2,050 |
| Payment submitted | 1,780 |
| Order confirmed | 1,720 |
Average order value: £38. Delivery is £3.95, free on orders over £50.
What the walkthrough and research found
| Step | Observation | Source |
|---|---|---|
| Product page | No statement on dye lots; several emails a week ask whether balls in one order will match | Support inbox |
| Product page | "Dispatched in 2 to 5 working days"; no delivery cost or arrival date | Walkthrough |
| Basket | "Delivery calculated at checkout"; free-delivery threshold not mentioned | Walkthrough |
| Basket | Open discount code box above the checkout button; shoppers leave and return from coupon sites | Recordings |
| Account | Account required; guest option is small grey text under the password form | Walkthrough, recordings |
| Account | Most poll answers say "didn't want another account" | Exit poll |
| Delivery | Delivery charge and threshold appear here for the first time | Walkthrough |
| Delivery | Seven manual address fields, no postcode lookup; phone number required with no reason given | Walkthrough, recordings of mobile shoppers zooming |
| Payment | Cards only, no wallets; redirect to a provider page with different branding | Walkthrough |
| Payment | A declined card clears the form | Support inbox |
Developer's estimates: guest checkout by default, half a day (a platform setting); delivery cost, date and threshold on product page and basket, 2 days; collapsed discount link, half a day; postcode lookup, autocomplete and a phone reason, 3 days; wallets and embedded card fields, 5 days plus a change to the payment contract; dye lot statement, 1 day of copy once the warehouse confirms its picking policy.
Recording friction
For each step, record four things: the shoppers lost there, what the shopper experiences, the evidence and how many independent sources support it, and the fix with its effort. Measure each step as its own rate, as the checkout guide insists, because a single checkout CVRShare of visits or clicks that convert. hides which step leaks. Note friction that starts upstream: a doubt created on the product page can show up as a loss at the basket.
Task
- Calculate the step rate and the number of shoppers lost at each step of the FunnelModel of the customer journey stages..
- Map every observation to its step and to the friction point it represents.
- For each friction point, state the fix, its effort and the strength of its evidence.
- Rank the fixes, explaining how you traded off loss, evidence and effort.
- Decide which fixes to test and which to ship directly, and which tests must run in sequence.
- Name the guardrails for the first test.
Worked answer
Step rates and losses
| Step | Rate | Lost |
|---|---|---|
| Basket to checkout start | 4,100 ÷ 7,800 = 52.6% | 3,700 |
| Account step | 2,700 ÷ 4,100 = 65.9% | 1,400 |
| Delivery step | 2,050 ÷ 2,700 = 75.9% | 650 |
| Payment step | 1,780 ÷ 2,050 = 86.8% | 270 |
| Confirmation | 1,720 ÷ 1,780 = 96.6% | 60 |
The average basket of £38 sits £12 below the free-delivery threshold.
The basket loses the most people, but many basket viewers are still browsing; inside the checkout, the account step is the largest leak by far, then delivery.
Ranked fixes.
- First: guest checkout by default. The biggest checkout leak, three sources agree, and the build is trivial. The checkout guide's fix is to offer the account after the order, on the confirmation page.
- Second: delivery cost, arrival date and the threshold on the product page and basket. It attacks the delivery step and part of the basket loss, and the evidence is the walkthrough plus the delivery step's drop. Expect some shoppers to leave at the basket instead of later, so judge it on orders per basket view, not on checkout starts. Showing the gap to free delivery may also lift AOVAverage revenue per order., which makes order value a measure to watch.
- Third: postcode lookup, autocomplete and a reason for the phone number. It targets the rest of the delivery loss, with recordings behind it and a modest build.
- Fourth: payment wallets, a consistent payment page and errors that keep the input. Real friction, but the smallest loss, the biggest build and a contract dependency, so it is recorded with that dependency and scheduled later.
Ship without a test. The collapsed discount link and the dye lot statement are low-risk, low-effort answers to documented problems. The dye lot line is exactly the kind of specific answer the product page guide asks for near the button, and it waits only on the warehouse confirming what can be promised. Clearing the card form on a decline is a bug; fix it.
Test and sequence. Guest checkout and the delivery changes touch consecutive steps, so they run one after the other, guest checkout first. Its primary metric is the account step rate, with orders per checkout start, order value and the share of shoppers creating an account after ordering as guardrails. The experimentation programme playbook covers how both tests enter the roadmap with owners and a read-out.
Common mistakes
- Treating the checkout as one rate. The leak is at a step, and only step rates show which.
- Judging the delivery-cost change on checkout starts. Fewer, better-informed starters can mean more orders.
- Testing bug fixes. A form that clears on error is broken; fix it and watch the step rate.
- Ignoring the product page. Doubts about dye lots and delivery begin there and surface at the basket.
Next, bring everything from this level together in the experimentation roadmap capstone, or return to the intermediate CRO level.
