UK ecommerce benchmarks by industry: what the August 2026 data shows, and what it cannot tell you

UK ecommerce benchmarks for August 2026 put the cross-market session conversion rate at 2.23% and average order value at £129.23, but the vertical spread runs from 0.57% to 5.81% — which means a cross-market average is the least useful number on this page. Two things matter more than the benchmark itself: this panel groups organic search into “Direct,” so it says nothing about SEO performance, and conversion rate on its own cannot tell you whether a store is healthy. Revenue per session can.

The Problem With How Benchmarks Get Used

A merchant reads that the average ecommerce conversion rate is around 2%, checks their own at 1.4%, and concludes something is broken. Or reads it, sees 3.1%, and concludes nothing needs attention.

Both readings are usually wrong, and for the same reason: conversion rate is a ratio whose denominator is traffic you chose to buy. A store running 80% paid social will look poor against one running 80% email, regardless of how well either is operated. The benchmark measured a different traffic mix than yours.

The vertical spread in this month’s data makes the point better than any argument. Arts & Crafts converts at 5.81%. Baby & Child converts at 0.57% — and carries an average order value of £855.57. Ten times the conversion rate, one-seventh the order value. Neither store is performing badly. They are selling different things to people making different decisions.

What the August 2026 Data Shows

Across the whole panel: session conversion rate 2.23%, up from 1.85% a year earlier. Average order value £129.23, up 6.28%. Revenue per session £1.92, up 24.65%. Cost per acquisition 9.17% of revenue, up from 8.60%. Bounce rate 39.75%, essentially flat. Sales up 7.65% year on year.

And the number that reframes all of the above: visitors down 13.69%.

By Vertical:

Device: Mobile 63.7% of sales, Desktop 34.7%, Tablet 1.6%.

Channel share of sales: paid search 64.5%, direct 19.3%, email 8.7%, affiliate 5%, other 2%, paid social 0.4%. International sales 41.3% of the total.

The Three Things This Data Cannot Tell You

This section matters more than the tables, and it is the part most benchmark articles omit.

1. It says nothing about organic search. IRP’s attribution model groups free traffic — direct visits, organic search, and unpaid referrals — together under “Direct.” That 19.3% is not an organic search figure and cannot be read as one. Any article that cites this dataset to make a claim about SEO performance has misread it. If you need organic benchmarks, this is the wrong panel.
2. The scope is narrower than “ecommerce.” The Market Data Centre covers B2C ecommerce in Great Britain, Northern Ireland, and Ireland, weighted toward independent SME and mid-market merchants — IRP uses roughly 30% as a working estimate of the market held by independents rather than by Amazon and the major retailers. It is not global, not enterprise, and not marketplace. A US DTC brand on Shopify is not in this population.
3. It cannot tell you whether you are doing well. Conversion rate is transactions divided by sessions, on last-click attribution. Change your channel mix and the number moves without anything about your store changing. The benchmark tells you what a differently composed group of merchants recorded, not what you should record.

There is a fourth worth noting: different panels disagree. Published benchmark sources have been found to differ by close to a factor of two on cross-industry averages, and to contradict each other on whether mobile or desktop converts better. Pick one panel, state which, and stay with it. Blending several into a single table produces a dataset that does not exist.

Who These Numbers Are Useful To

If you sell B2C in the UK or Ireland, run an independent or mid-market operation, and want direction of travel rather than a target, this panel is well matched to you and the monthly cadence is genuinely useful.

If you sell internationally, operate at enterprise scale, run primarily through marketplaces, or sell B2B, treat these as context and find a panel that matches your population. And if the question you are answering is about organic search, this dataset cannot answer it at all.

The Story in This Month’s Numbers

Three patterns are worth more than any individual figure.

Traffic is down and revenue is up. Visitors fell 13.69% while sales rose 7.65% and revenue per session rose 24.65%. Fewer people, each worth considerably more. That combination has been building through 2026 and it changes what “good” looks like — a store reporting falling sessions and rising revenue is not in decline, it is tracking the market.

It is tempting to attribute the traffic decline to AI-driven zero-click behaviour, and the timing fits. This dataset cannot demonstrate that, because it does not separate organic search. Treat it as a hypothesis to test in your own analytics rather than a conclusion.

Acquisition is getting more expensive. Cost per acquisition rose to 9.17% of revenue and cost per session rose 36.96%. Revenue per session is outrunning it for now, but the gap is narrowing and this is the pressure to watch going into peak.

Mobile carries the majority of sales. 63.7% against desktop’s 34.7%. This contradicts a widely repeated claim that desktop dominates ecommerce revenue, and it is a reminder that “mobile browses, desktop buys” is a generalisation that depends entirely on which panel you read.

The Metric to Use Instead

Conversion rate is the most quoted ecommerce metric and among the least useful in isolation, because it ignores basket size entirely. Average order value has the opposite problem.

Revenue per session combines both. Across this panel it is £1.92, up 24.65% year on year. A store whose conversion rate falls while AOV rises may be performing better than it was; revenue per session tells you which, and conversion rate alone cannot.

Add cost per session — £0.15 here, up 36.96% — and you have the two numbers that actually describe whether traffic is worth buying. The gap between them is your margin on acquisition, and it is the figure that should drive spend decisions.

How to Compare Properly

• Match the population before the metric. Geography, business size, channel mix, B2C or B2B. If the panel does not match your store, the number is trivia.
• Check the definition. Conversion rate here is transactions divided by sessions. Others use orders divided by users, or purchases divided by visitors. These produce materially different numbers from identical trading.
• Check the attribution model. Last-click here, with free traffic bucketed into Direct. A panel using data-driven attribution will report a different channel split from the same underlying behaviour.
• Clean your own data first. Bot traffic and duplicate purchase events distort a store’s own conversion rate before benchmarking begins, and neither shows up as an obvious error. Establish that your number is real before comparing it to anyone’s.
• Compare to your own vertical, or not at all. With the spread running from 0.57% to 5.81%, a cross-market average is not a target for anybody.
• Track month over month, not month against benchmark. Direction of travel is where the signal is.

What Goes Wrong

• Blending panels into one table. Different samples, definitions and attribution models presented as a single dataset. This is the most common error in benchmark content and it is invisible to the reader.
• Citing session-based data to make an SEO claim. If organic is bucketed into Direct, the dataset cannot support the argument.
• Comparing to the cross-market average. It is a weighted blend of ten verticals that behave nothing like each other.
• Reading a conversion rate drop as a problem. Check AOV and revenue per session first; the mix may simply have changed.
• Ignoring traffic mix. A shift toward paid social will lower conversion rate with no operational change at all.
• Using a benchmark as a target. It describes a population you are not in.
• Quoting a figure without its date. This data updates monthly and this month’s numbers moved 20% year on year.
• Skipping the attribution footnote. It determines what the channel split means, and it is usually the last thing anyone reads.

Using the Table Without Being Misled

That last row is worth stating plainly: this panel does not publish Beauty or Electronics as separate verticals. If you need those, you need a different source — and you should say which one, rather than importing a figure from elsewhere into this table.

Shopify website development company