Conversion rate calculator
Conversion rate is conversions divided by the traffic that produced them, as a percentage. This panel divides by clicks: 100 conversions from 2,000 clicks is 5%. Change the denominator to sessions or to people and the same 100 conversions gives a different number, so the denominator has to travel with the figure.
How to calculate conversion rate
Conversion rate earns its reputation arithmetically. It divides into every upstream cost at once, so doubling it halves CPA and doubles ROAS without a single bid changing. On the figures above, a click costs 0.50 and a conversion costs 10; lift the rate from 5% to 6% and the conversion cost falls to 8.33 with the media plan untouched. Nothing else in the funnel pays back at that rate.
The denominator is where the number goes wrong, and this panel has made a choice on your behalf: conversions divided by clicks. An ad platform counts a click when someone taps the ad. Your analytics counts a session when the page loads and the tag fires. The two never match, because people leave before the page renders, because the platform and the analytics filter bots differently, and because one person clicking three ads is three clicks and often one session. Sessions in turn exceed people. Feed the same 100 conversions into all three and you get three defensible conversion rates, and comparing your clicks-based figure against somebody else’s sessions-based benchmark compares nothing at all.
The window is the second trap and gets far less attention. This is a ratio of two counts taken over the same dates, but the conversions in that period were not all produced by the clicks in it. Where the purchase decision takes a week, roughly a week of the conversions you are counting came from clicks before the window opened, while a week of the clicks inside it have not converted yet. If spend is flat the two errors cancel and nobody notices. If spend is climbing, the denominator inflates before the numerator catches up and the rate looks like it is falling. If you have just cut spend, the rate flatters you. Scaling a campaign and watching its conversion rate sag is very often this artefact rather than a real fall in quality.
For a benchmark with a stated definition rather than a remembered one: IRP Commerce publishes monthly e-commerce market data from its own platform, using transactions divided by sessions. In July 2026 its overall market rate was 2.26%. The useful part is not that figure but the spread underneath it: across the ten verticals it reported that month, Arts and Crafts converted at 5.23% and Baby and Child at 0.55%, a range of nearly ten to one. A single cross-industry number cannot tell you whether your rate is good, and any benchmark quoted without its category, its denominator and its date is decoration.
Two rows on this panel need reading with care. Profit here is revenue minus ad spend and nothing else, so it does not know about your cost of goods, your fulfilment or your fees; on the defaults it reads 4,000 on 5,000 of revenue, which is a number no shop has ever banked. Revenue per click is the figure to compare against your cost per click, and the pair of them replaces the profit row usefully: 2.50 in against 0.50 out.
Finally, the sample size. A 20% relative improvement measured on 200 visits is well inside the noise, and the temptation to act on it is strong precisely because the number moved a lot. Sizing a comparison between two variants is a different calculation from sizing a survey: it needs a baseline rate, the smallest difference you would act on, and a power level, and it produces a sample per arm. The sample size calculator on this site answers the survey question instead, sizing a single proportion to a chosen margin of error, so use it to ask how precisely you know one rate rather than to size a two-arm test. As a rough feel for the scale involved, distinguishing 5% from 6% reliably takes thousands of visits per variant, not hundreds.
What people use it for
- Measuring a landing page test
- Comparing traffic sources on quality rather than volume
- Forecasting revenue from a traffic target
- Diagnosing a drop in sales
- Reconciling an ad platform’s conversion rate with the one in your analytics
- Checking whether a rate that fell after scaling actually fell
- Deciding whether a difference between two variants is large enough to act on
Questions
It depends entirely on the category. IRP Commerce reported an overall e-commerce market rate of 2.26% for July 2026, with its verticals that month running from 5.23% down to 0.55%. Compare against your own history first.
Either, provided you are consistent and you say which. Ad platforms count clicks, site analytics counts sessions, and the two never agree. This panel uses clicks.
Different denominators, different attribution windows and different bot filtering. The platform counts a click it served; analytics counts a session where its tag fired. A gap of ten to twenty per cent between the two is ordinary.
For a considered purchase, often yes. Per-user conversion rate is higher than per-session, because most people take several sessions to buy, and it answers "how many of these people bought" rather than "how many of these visits ended in a sale".
Frequently because the numerator lags the denominator. New clicks arrive immediately; their conversions arrive over the following days or weeks, so a rising spend depresses the ratio before any real change in quality.
Compare cohorts rather than calendar periods: take the clicks from one week and count the conversions they eventually produced, however late. Most platforms offer this as a conversion-date versus click-date setting.
Yes, the numerator moves with it. Last-click credits the final touch; a data-driven model spreads a conversion across several. The same sales can raise or lower a given channel’s rate without a single customer behaving differently.
Because it divides into every upstream cost. On the figures above, moving 5% to 6% takes cost per conversion from 10 to 8.33 with no change to bids, creative or budget.
Track them separately. Folding newsletter signups into a purchase rate makes the headline look better and stops it meaning anything, and it hides the fact that the two respond to completely different changes.
Whatever you have marked as a key event. That is a configuration choice rather than a fact, so two properties on the same site can report different conversion rates entirely legitimately.
More than intuition suggests. Telling 5% from 6% at conventional confidence runs to thousands of visits per variant. Hundreds of visits can move a rate by a fifth on noise alone.
Not for a two-variant test. It sizes a single proportion to a margin of error, which is the survey question. A test between two variants needs a baseline, a minimum detectable effect and a power level, and gives a sample per arm.
No. Tightening targeting until only ready-to-buy traffic arrives raises the rate and can shrink total sales. Rate is a ratio, and the business runs on the numerator.
Because the intent arrived before the click did. Mixing branded and non-branded traffic into one rate hides both, and a rise in the blend usually reflects a change in traffic mix rather than in the site.
Substantially. Mobile typically converts well below desktop on the same site while carrying more of the traffic, so a blended rate mostly tracks your device mix. Split it before drawing conclusions.
They convert far better than new ones and are not evidence that anything is working. Segment new from returning before comparing periods, or a retargeting campaign will appear to fix your whole funnel.
No. It is revenue minus ad spend, with no cost of goods, fulfilment or platform fees in it. Compare revenue per click against your cost per click instead.
Inversely and directly: CPA is cost per click divided by conversion rate. Halving the rate doubles the CPA at the same bid.
Volume, with rate as the diagnostic. Rate tells you where the funnel leaks; it does not tell you the leak is the biggest opportunity available.
Only with care. Seasonality, promotions and traffic mix all move the rate without anything on the site changing. Year on year for the same month is usually the cleaner comparison.
No. The figures stay in the page and nothing is uploaded.