Measure the checkout step by step
Opinions about checkouts are plentiful. Everyone has a theory: too many fields, the wrong button colour, not enough payment logos. The checkout itself will tell you what is wrong if you count. Set up measurement so that each step is recorded, and look at the numbers split in a few ways.
| Step | What to count | Split by |
|---|---|---|
| Basket viewed | Sessions that open the basket | Device, new or returning |
| Checkout started | Sessions that begin checkout | Device, signed in or guest |
| Contact and address completed | Sessions that pass the details step | Device, country |
| Delivery method chosen | Sessions that pass delivery | Delivery option, basket value |
| Payment attempted | Sessions that submit payment | Payment method |
| Payment succeeded | Orders placed | Payment method, card type, country |
| Errors shown | Every validation and system error, with its text | Step, field, device |
The largest drop between two adjacent steps is where to look first. A sharp difference between phone and desktop, or between one payment method and the others, usually points straight at the cause. Keep personal details out of analytics tools when setting this up.
Find out why shoppers leave
- 01
Place test orders yourself
On a phone, on mobile data, as a new customer, with a real card, to a real address. Then repeat with each payment method and delivery option. Note every hesitation.
- 02
Try to break it
A long name, an apartment address, a postcode with a space, a foreign phone number, an expired card, the back button halfway through.
- 03
Watch recorded sessions
Session recording tools show where people pause, retype and give up. Watch twenty sessions that abandoned at your worst step.
- 04
Read what customers say
Support messages, chat transcripts and reviews that mention ordering or payment. Ask your support team what they hear.
- 05
Check the error log
The errors counted in the table above, read as text, often name the fault outright.
- 06
Ask a few shoppers
A one-question survey on leaving the checkout, or calls to customers who abandoned, gives reasons that data cannot.
Some abandonment is not a fault. People use the basket to check delivery costs, save items or compare prices. The aim is to remove obstacles for those who intended to buy, and no checkout converts everyone.
Common causes and what to do about them
| Cause | How it shows | Usual remedy |
|---|---|---|
| Costs revealed late | Drop at the delivery or payment step | Show delivery cost, taxes and fees in the basket or on the product page |
| Forced account creation | Drop at the start of checkout among new customers | Offer guest checkout, and invite an account after the order |
| Too many or confusing fields | Slow completion and errors on the details step | Remove optional fields, use address lookup, label clearly |
| Poor phone experience | Phone completion far below desktop | Correct keyboards for each field, large targets, wallet payments |
| Missing payment methods | Drop at payment, differing by country | Add the methods customers in each market expect |
| Payment failures | Attempts well above successes | Clear decline messages, easy retry, a second method, review with the provider |
| Unclear errors | Repeated attempts on one field | Say what is wrong and how to fix it, next to the field, keeping what was typed |
| Slow or unstable pages | Drop everywhere, worse on phones | Performance work on the basket and checkout |
| Doubts about trust | Hesitation before payment | Visible returns policy, delivery times, contact details and security cues |
Rank fixes by effect and effort
List the causes you found, and for each estimate how many checkout sessions it affects and how hard it is to fix. A fault that touches every phone shopper outranks one that touches a rare card type, even if the second is more annoying to the person who reported it.
Know your platform's limits before promising anything. Hosted ecommerce platforms control much of the checkout for security and reliability, and how far it can be changed depends on the platform and often on the plan. Settings, payment methods, delivery rules and text are usually within reach. Restructuring the steps may not be. Where the checkout is your own code, everything can be changed, and every change needs testing against payments and fraud controls. Our comparison of headless and traditional ecommerce explains what that freedom costs.
Look closely at payments
- 01
What share of payment attempts succeed?
What to checkBy method, card type and country. A low rate for one group points to a configuration or provider problem.
- 02
What do declined customers see?
What to checkA message that says try another card or contact your bank, with the basket intact, recovers orders. A generic failure loses them.
- 03
Is extra authentication handled well?
What to checkBank verification steps add friction and can fail on phones. Test them on real devices and ask your provider what can be tuned.
- 04
Are wallets offered?
What to checkApple Pay, Google Pay and similar methods remove typing altogether on phones.
- 05
Are local methods offered?
What to checkPreferred methods differ by country: bank transfer schemes, instalments, cash on delivery. Check what your customers in each market expect.
- 06
Are fraud rules too strict?
What to checkRules that block genuine customers cost sales quietly. Review blocked orders with your provider.
Your payment provider holds data you cannot see in the store's reports, including decline reasons. Ask for a review. Our payment gateway integration page covers how payment is connected and tested.
A worked example
A homeware store reviews its checkout
This example is invented to show evidence overturning assumptions. A homeware store's owner believes the checkout has too many steps and wants it rebuilt as a single page.
- MeasurementStep counts show the largest drop at delivery, not at the details step. On phones, payment attempts exceed successes by a wide margin for one card type.
- WhyTest orders reveal that delivery cost appears only at the third step and is high for large items. Session recordings show shoppers leaving within seconds of seeing it.
- PaymentThe provider confirms that bank verification for that card type fails in the store's embedded payment form on some phones. A setting change routes it through a full-page step.
- ChangesA delivery estimate is added to product pages and the basket. The payment setting is changed. The number of steps stays as it was.
- ResultOver the following month, completion from basket to order rises clearly on phones, and payment success for the affected card type matches the others.
A single-page rebuild would have cost far more and fixed neither problem. Both causes were found in a week of measuring and testing, and both fixes were small.
Test changes and keep what works
Sound testing
- One change at a time where traffic allows
- A split test when volumes are high enough to trust the result
- Before and after compared over similar periods when they are not
- Success measured as completed orders, not clicks
- Every change tested with real payments on real phones first
Unsound testing
- Five changes released together
- Judging by a few days of data
- Comparing a sale week with a normal week
- Declaring success on a measure nobody defined beforehand
- Changing the checkout just before peak season
Small stores rarely have enough orders for a split test to reach a reliable result quickly. That is no reason not to improve. Fix clear faults, compare like periods, and give more weight to large, obvious effects than to small differences.
Checkout improvement checklist
Checkout improvement checklist
0 of 11 checkedWhat to do next
This evening, buy something from your own store on your phone, using mobile data and a card you have not used there before. Write down everything that slowed you. Then look up the count of sessions at each step for last month. Those two things usually set the plan.
If the findings point beyond settings, our e-commerce platform development page explains how we work on checkouts, and a UX audit gives the whole purchase journey an independent review. Tell us what your numbers show.
Common questions
Is a one-page checkout better than several steps?
Not reliably. What matters is how much effort the shopper must spend and whether each step is clear. A well-made checkout in several short steps can outperform a long single page, and the reverse. Measure where people leave in your own checkout. The number of pages is rarely the cause, and rebuilding for its sake seldom pays.
Should we require customers to create an account?
Requiring it before a first purchase puts a task between the shopper and the order, and new customers often leave. Offer guest checkout, and invite the customer to save their details with a password after the order is placed. Returning customers benefit from accounts, so make signing in easy without making it compulsory.
How many payment methods should we offer?
Enough to cover what your customers in each market expect, and no more. Cards and the main digital wallets are the usual base. Add local methods where a meaningful share of customers prefer them. Too many options clutter the page, so show the relevant ones for the shopper's country and device first.
Why do payments fail when the card is valid?
A bank may decline for its own risk reasons, extra verification may fail or time out, details may be mistyped, or your own fraud rules may block the order. Your payment provider can show decline reasons by category. Clear messages, an easy retry and an alternative method recover many of these orders.
How much can we change on a hosted platform's checkout?
It varies by platform and plan. Branding, text, fields within limits, payment methods and delivery options are generally adjustable. The structure and code of the checkout are often controlled by the platform. Check current documentation for your plan before planning changes, and ask whether an extension mechanism exists for what you need.
