Conversion rate is the percentage of visits, or of people, that complete the action a page exists for: an order, a submitted form, a phone call, a quote request.
The first thing to know: the number you are looking at depends on what sits in the denominator. Google publishes three different ones, with three different formulas, and the results cannot be compared with each other. If you put your 1.8% next to a 2.23% benchmark without knowing which of the three you are using, you are comparing two different measurements.
How to calculate conversion rate
Conversions divided by the total, multiplied by one hundred. With 4,000 sessions and 72 orders the sum is 72 ÷ 4,000 × 100 = 1.8%.
The Google Ads documentation is more precise than most of what you read online. The denominator is not clicks: it is the ad interactions that can be tracked to a conversion in the same period. Google's own example is 50 conversions from 1,000 interactions, which gives 5%.
The same page carries a warning that surprises people the first time they read it: the rate can go above 100%. It happens when you track several conversion actions on the same campaign, or when counting is set to "every" conversion rather than "one". A user who downloads two PDFs and fills in a form counts as three conversions and a single interaction.
GA4 has no conversion rate. It has two
Since 21 March 2024 Google Analytics no longer calls the events that matter to the business conversions: they are key events. The term conversion stays in Google Ads, and Google's blog announcement explains that the split exists precisely to keep the two measurements apart.
The practical consequence affects anyone who opens a report. In the official table of Analytics dimensions and metrics there are two entries, not one.
- Session key event rate: the number of sessions in which a key event occurred, divided by the total number of sessions.
- User key event rate: the number of users who triggered a key event, divided by the total number of users.
A user who comes back four times before buying is worth one conversion across four sessions (25%) or one conversion from one user (100%). Same purchase, two numbers far apart. In the GA4 API the two metrics have separate names, sessionKeyEventRate and userKeyEventRate, and the developer documentation states that there are two rate metrics for every key event.
| Metric | Denominator | Where you find it | API name |
|---|---|---|---|
| Session key event rate | Total sessions | GA4 | sessionKeyEventRate |
| User key event rate | Total users | GA4 | userKeyEventRate |
| Conversion rate | Ad interactions that can be tracked to a conversion | Google Ads | conversion rate |
This is also why the GA4 figure almost never matches the one in your e-commerce back end. It is not a tracking problem: the denominators are different. The same logic applies to bounce rate, which changed definition in GA4 without changing its name.
What is a good conversion rate
It depends on the sector and on the traffic mix, but there are three reference points with a stated method and market.
IRP Commerce publishes monthly trading data for the merchants on its platform, calculated as transactions divided by sessions. In August 2026 the average conversion rate was 2.23%, up from 1.85% in August 2025. By sector the values spread widely: Arts & Crafts 5.81%, Health & Wellbeing 3.31%, Fashion 1.86%, Food & Drink 1.58%, Baby & Child 0.57%. The sample is British and Irish, and the page is updated every month.
WordStream, part of LocaliQ, measures search campaigns instead: 13,474 US campaigns between April 2025 and March 2026, Google Ads and Microsoft Ads combined, with a median of 8.18%. It looks huge next to a website, and the reason is that paid traffic on commercial queries arrives already filtered. You pay for that filter, and the price shows up in the cost per click. For comparison, in the first report of the same series, in 2016, the median was 2.70%.
Unbounce analyses 41,000 landing pages with 464 million visits and reports a median of 6.6% for Q4 2024. The market is not stated, and the sample is made of pages built specifically to convert: it is useful for calibrating a B2B landing page, not a whole website.
| Source | What it measures | Value | Market | Period |
|---|---|---|---|---|
| IRP Commerce | Transactions / sessions, e-commerce | 2.23% | UK and Ireland | August 2026 |
| WordStream (LocaliQ) | Google and Microsoft Ads search campaigns | 8.18% (median) | United States | Apr 2025 - Mar 2026 |
| Unbounce | Landing pages | 6.6% (median) | not stated | Q4 2024 |
A UK benchmark exists, but it is not yours
Search for "average conversion rate UK" and you will find a handful of percentages that have been passed around for years, often with no primary source behind them. The IRP Commerce figure is one of the few with a published method, and it is British, but it is an average across merchants on a single platform, with their own sector mix and price points.
The context around it comes from the Office for National Statistics: in August 2026 internet sales were 27.3% of total UK retail sales, against 26.1% a year earlier, with a peak of 32.4% in November 2025. The share of shopping done online moves with the calendar, and so do conversion rates. A monthly average taken in a Black Friday month and one taken in August are not describing the same market.
The benchmark that really counts is your own. Take twelve months, separate brand traffic from non-brand traffic, keep mobile and desktop apart, and compare each segment with itself month on month. If you work in lead generation, add the next stage, the lead-to-customer rate: a form that converts twice as well and brings in leads that never close has made the funnel worse, not better. The same reasoning applies to every SEO KPI.
What actually moves conversion rate
The levers listed in every guide are real: loading speed, a clear offer, shorter forms, a checkout with fewer steps, reassurance shown at the right moment. Core Web Vitals measure the first of these and have public thresholds, so they are the easiest place to start.
What you read less often is that the make-up of your traffic matters before the page does. A navigational query with the brand name and a generic informational query bring two audiences at very different distances from a purchase, and averaging them into one rate hides both. Segmenting by search intent before optimising anything stops you working on a page that is already doing its job.
With A/B tests there is also a sum that hardly anyone does before launching them. The rule of thumb for sizing a two-variant test is n ≈ 16 × p × (1 - p) / δ², where p is the current rate and δ the absolute difference you want to be able to detect. With a rate of 1.5% and the aim of detecting a 20% relative improvement, that is 0.3 percentage points, you need about 29,000 sessions per variant. A site with 10,000 sessions a month would take almost six months to finish a single test, seasonality included. Below that level of traffic, tests on the shade of a button give no answers: you are better off moving to offer, price and audience, where the effects are large enough to show up even with little data.
Conversion rate FAQs
Conversions divided by the reference total, multiplied by one hundred. In Google Ads the denominator is the ad interactions that can be tracked to a conversion: 50 conversions from 1,000 interactions make 5%. In GA4 there are two possible denominators, total sessions or total users, and they return two different values for the same event.
There is no value that holds everywhere, because it depends on the sector and the type of traffic. The reference points with a stated method are IRP Commerce's 2.23% for UK and Irish e-commerce in August 2026, WordStream's 8.18% median for US search campaigns and Unbounce's 6.6% median for landing pages in Q4 2024. The benchmark that matters most is your own, segmented and compared month on month.
Because they measure different things. GA4 divides key events by sessions or by users collected in the browser, while the platform divides the orders recorded in its system. Different denominators and different collection methods produce different percentages even when the number of orders is identical.
In Google Ads, yes. The official help page explains that it happens when several conversion actions are tracked or when counting is set to every conversion rather than one: the same interaction can generate more than one conversion.
The session rate is for assessing individual pages, campaigns and short periods. The user rate is for long decision cycles, where people come back several times before deciding, typically in B2B and on high-value purchases. Google publishes both precisely because they answer different questions.
One thing the industry rarely says: conversion rate is a fraction, and in most cases it is the denominator that moves. An SEO campaign that works brings in informational traffic, total sessions grow and the percentage falls while orders rise. Anyone who looks only at the percentage sees a decline where there is growth. Before starting an optimisation project, it is worth checking whether the numerator, in absolute terms, has moved in the opposite direction.
