An online shop is judged on one thing: how often a visit becomes an order. Conversion rate is the most quoted figure in ecommerce and probably the worst defined. Depending on which dashboard you open, two people can describe the same shop with very different numbers, and neither is wrong.
Ecommerce conversion rate optimisation begins with that definition problem, not with a redesign. What follows sets out definitions and scope, gives published benchmarks with their limits, then works through the documented levers: checkout flow, cost visibility, payment methods, speed, proof and returns. It closes on the legal layer, which is not administrative trivia — it dictates part of your checkout.
Define the rate before you argue about it
The most common formula is the session conversion rate: transactions divided by sessions, multiplied by 100. That is the definition IRP Commerce publishes in its market data centre, under the heading "Session Conversion Rate". The other convention divides orders by unique visitors over a period. Because one visitor often generates several sessions before buying, the per-visitor rate is mechanically higher. That is not better performance. It is a smaller denominator.
Scope matters as much as the formula. Are you counting every session, including those landing on a blog post? Or only sessions that reached a product page? Are bots and internal traffic filtered out? In audit work, the gap between two dashboards for one shop almost always comes down to those choices, not to a bug.
| Formula | What it measures | Where it fits |
|---|---|---|
| Orders ÷ sessions | How productive a single visit is | Daily monitoring, channel comparison |
| Orders ÷ unique visitors | The share of buyers in your audience | Monthly reporting, cohort analysis |
| Orders ÷ sessions with a product view | How well the buying path itself works | Funnel diagnosis, editorial traffic excluded |
Pick one and do not change it without saying so. A rate that "improves" because the denominator moved is the most common confusion in any steering meeting. Write the definition down:
Conversion rate — recorded definition
Numerator : completed orders, back office
Denominator : sessions, analytics tool, all devices
Scope : whole site, bots and internal traffic excludedWhat to compare your rate against
Sector benchmarks exist, but each has a narrow perimeter. IRP Commerce, whose data covers UK and Irish ecommerce and the merchants on its own platform, recorded a session conversion rate of 2.26% in July 2026, against 1.94% in July 2025. The 2026 Digital Experience Benchmark from Contentsquare, built on more than 6,500 sites and 99 billion web and app sessions, measured conversion down 5.1% between the fourth quarter of 2024 and the fourth quarter of 2025, with traffic down 3.8%.
One French figure is worth borrowing for the shape of the trend, read as French data and not as a European average. In its review of 11 February 2026, the French ecommerce federation FEVAD put the 2025 French market at 196.4 billion euros, up 7% year on year, across 3.2 billion transactions (+10%), with an average basket of 62 euros, down 3%. Growth came from purchase frequency, not order value.
Benchmarks give an order of magnitude, not a target. A spare-parts shop, a fashion shop and a ticketing site have no reason to converge. Your only solid reference is your own historical series, at constant scope.
Consent shapes what you are able to measure
The number you read depends on what your visitors agreed to. Across the UK and the EU, non-essential tracking requires consent, and analytics scripts usually fall on the wrong side of that line unless they are tightly restricted.
The French regulator publishes the strictest version of that test. On its guidance page updated on 4 July 2025, the CNIL exempts audience-measurement trackers from consent only where measurement serves the publisher alone, the statistics are anonymous, nothing is cross-referenced or passed to third parties, there is no tracking across sites, and the tracker lifetime is capped at thirteen months. That is French guidance, but it makes a useful checklist anywhere: common analytics and advertising tools rarely meet it.
So the transaction count in your analytics tool is almost always lower than the one in your back office. The method that holds up: back-office orders in the numerator, analytics sessions in the denominator, the gap documented, the same rule every month. That plumbing belongs in ongoing traffic management, not in a one-off report.
The levers that are actually documented
Cart abandonment is massive and structural. Baymard Institute, which aggregates fifty studies on the subject, calculates an average abandonment rate of 70.22% (list updated 22 September 2025). Part of that is irreducible: in Baymard's survey of US online shoppers, 42% said they had abandoned a cart because they were "just browsing". Among the other stated reasons in that US survey, 40% cited extra costs that were too high, 19% distrust about entering card details, 18% forced account creation and 17% a checkout that was too long.
Show the full cost as early as you can
The heaviest actionable reasons involve money discovered too late; a further 12% in the same survey could not see the total cost up front. A shipping estimator on the product page, a legible free-delivery threshold, a total calculated before the address form. None of this makes the order cheaper. It removes the surprise.
Shorten the checkout and allow guest orders
Baymard puts a well-designed checkout at 12 to 14 form elements, which is 7 to 8 input fields, against an average of 23.48 across the US sites in its benchmark. Remove an optional field, pre-fill the town from the postcode, offer a guest checkout. A UX audit exists to rank those frictions rather than tackle them at random.
Widen payment methods, with discernment
In the same US survey, 9% of stated abandonments came down to insufficient payment methods. But every extra provider carries commission and maintenance costs. Add what your customers ask for, measure the share of orders it captures, then decide.
Speed, measured separately on mobile
The thresholds are public and stable. According to Google's Core Web Vitals documentation (updated 31 October 2024), a page counts as good below 2.5 seconds of LCP, 200 milliseconds of INP and 0.1 of CLS, at the 75th percentile and separately on mobile and desktop. That last point decides a lot: an all-devices average hides a slower mobile checkout. Working on site speed before a visual redesign usually returns more.
Proof and returns
Reviews and the returns policy do the same job: they reduce perceived risk. A clear returns page, reachable from the product sheet and written without jargon, is often among the most visited pages in a shop. Treat it as a conversion asset.
Checkout law is part of the funnel
Several obligations shape the final steps of an order, and they differ by market. If you sell into the EU, the baseline comes from EU consumer law: the final button must make clear that placing the order carries an obligation to pay, the order must pass through a correctable summary, and distance sales carry a fourteen-day withdrawal period with a defined list of exceptions.
Reviews and price claims sit in the same rules. Directive (EU) 2019/2161 of 27 November 2019 requires a trader publishing consumer reviews to state whether, and how, it checks that they come from real buyers; publishing or commissioning fake reviews is an unfair commercial practice. The same text frames price-reduction announcements: the reference price shown must be the lowest applied during at least the thirty days before the promotion.
If you trade in the UK, the equivalent duties sit in UK consumer legislation, not in the EU texts cited here. Check the exact wording with your own legal adviser before rewriting checkout copy.
Where to start
- Write down your definition of the conversion rate and its scope, then freeze both.
- Reconcile back-office orders against analytics orders monthly, and record the gap.
- Walk the funnel step by step, separately on mobile and on desktop.
- Deal with total-cost visibility first, then checkout length, then payment methods.
- Check the order button wording, the cancellation information and the review-verification statement.
- Change one thing at a time, and let enough orders pass for the difference to be readable.
One last point: conversion is also decided upstream. Badly targeted traffic drags the rate down without any page being at fault. If the structure of your shop no longer holds, work on the site itself comes before fine-grained funnel work.
Common questions
What counts as a good ecommerce conversion rate?
There is no universal value: the rate depends on the sector, price point, season and traffic quality. As a marker, IRP Commerce measured 2.26% per session in July 2026 across its UK and Irish panel. The reference that matters is your own history, on the same formula and scope each month.
Should I calculate conversion per session or per visitor?
Both are valid, but they do not produce the same number: a visitor often generates several sessions before buying, so the per-visitor rate is always higher. Per session suits daily monitoring and channel comparison; per visitor suits monthly reporting and cohort analysis. Do not mix them in one table.
Why does my analytics tool report fewer orders than my back office?
Because measurement depends on tracking consent: visitors who refuse non-exempt cookies are not counted by conventional analytics tools. Add script blocking, tagging errors, and orders placed by phone or in store. The back office is the source of truth for revenue; analytics explains the journeys behind it.
How much cart abandonment can realistically be recovered?
Not all of it. In Baymard Institute's survey of US online shoppers, 42% of abandonments were attributed to browsing with no intent to buy, which no interface change recovers. The actionable share sits in the other reasons — costs discovered late, forced accounts, a long checkout.
Does showing delivery costs earlier really change anything?
It is the most frequently cited actionable reason for abandonment. In the Baymard survey of US shoppers, 40% of those who abandoned pointed to extra costs that were too high, and 12% to not seeing the total up front. An early estimate does not reduce the price, but it removes the surprise at payment.
How long should I wait before judging a checkout change?
Long enough for the order volume to make the difference readable, which depends on your traffic, not on a fixed number of days. Change one thing at a time, and decide in advance what difference would count as meaningful.