What Is a Good Conversion Rate? Check What You Divide By

Claim: There Is an Average Conversion Rate You Can Measure Yourself By

A conversion rate is the share of people who did what you wanted, out of the people who had a chance to do it. Converted divided by opportunity, times a hundred. Thirty leads, six of them clients: that is 20%. Nobody struggles with that part.

“People who had a chance” is where it falls apart. It can mean everyone who landed on the site, everyone who reached checkout, everyone who filled in a form, everyone who clicked an advert, or everyone who saw one landing page. Same business, five different answers.

Here is what the widely quoted sources are each dividing by:

Almost seven percentage points separate the top from the bottom, and every entry is defensible. They are not measuring the same thing, and most articles that quote them never say so.

Claim: The Conversion Rate Benchmarks on Page One of Google Are Current

The most repeated figures on this topic, “median 2.35%, top ten percent 11.45%”, come from a WordStream article first published in March of 2014. The page shows an update in May of 2026.

According to its methodology, hundreds of accounts were examined over a three-month period after accounts with less than ten conversions or 100 clicks per month were eliminated. What is missing is when. No collection dates appear anywhere on the page, so a reader can’t tell whether the data behind the 2026 update is from 2026 or from 2014.

IRP Commerce does better on every count: formula printed on the page, coverage stated as British, Northern Irish and Irish B2C, a named month, and a monthly refresh. Ruler Analytics reports a big sample, 110 million sessions and 5 million conversions, and counts a conversion as a qualified lead or a sale, but publishes no formula and never names a country. Read its 5.13% as a lead-or-sale figure of unstated origin.

Claim: A Falling Conversion Rate Means Something Broke

Four things move this number without anything being wrong.

Your blog sits in the denominator. Publish more articles, pull in readers who were never buying today, and a session-based percentage drops while the content does its job.

Your traffic mix changed. Switch on paid social and colder visitors arrive in volume. The denominator grows faster than the numerator, so sales can rise while the percentage falls.

You are holding an advert rate against a site rate. That 8.18% divides leads by clicks from people who typed a buying phrase into Google. Your dashboard counts bots, blog readers and the same person visiting three times. The two do not compare.

Your reporting window is shorter than your sales cycle. Close in sixty days, measure in thirty, and the denominator is full of people who have not yet had the chance to say yes.

Claim: American Benchmarks Are Close Enough for a European Business

Almost every benchmark in an English-language search is American or unlabelled.

Landmark Global, citing the European E-commerce Report 2024, and Eightx’s 2026 EU benchmark land within a rounding error of each other on the Netherlands. So a Dutch shop at 1.8% sits on its national median, reads the American articles, and concludes it is failing.

Eightx describes the damage precisely: “When you drop a 2.5% benchmark on a French store converting at 1.2%, the store looks like it is leaving half its revenue on the table, and the natural reaction is to slash ad spend or rip apart a funnel that is actually performing at market.”

They also concede something more useful than what they publish. Their scorecard is stitched together, they write, “because no public dataset reports harmonized EU conversion, AOV, and returns by country and category for 2026.” The benchmark a Dutch or Belgian owner wants does not exist. Knowing that beats borrowing an American one.

Claim: You Can Test Your Way Up

True at scale, and out of reach for most small businesses. Seer Interactive puts the working threshold at a few thousand monthly visits and around a hundred conversions per variation. Their worked example: a page taking a thousand visits a day at 2% needs 103 days and roughly 103,000 visitors before the result means anything.

At a thousand sessions a month that test is not available to you, at any speed. Advice written for businesses with fifty times your traffic will have you redesigning pages on evidence indistinguishable from noise.

Claim: Your Own History Beats Every Published Benchmark

One comparison survives everything above, and IRP Commerce shows what it looks like. IRP fixes its formula, transactions divided by sessions, and holds it steady, so its July 2026 reading of 2.26% can be set beside July 2025 at 1.94% and the movement carries information. Same formula, same population, same month of the year. Nothing else on this page offers that.

You can build the equivalent in a spreadsheet, with no tools and no budget. Write the denominator in words at the top. Then one row per month: the month, the numerator, the denominator, the percentage. Twelve rows later you have a benchmark nobody can argue with, because it came out of your business rather than out of somebody else’s sample.

It also answers the two claims above it. Holding the definition fixed is what lets you see whether a drop is real, and unlike an A/B test it stays valid at any volume.