How to calculate conversion rate (with worked examples)

The formula, and the choice that changes the answer

The formula is simple: conversion rate = conversions ÷ visits × 100. Google Ads Help gives the example of 50 conversions from 1,000 interactions, which is 5%. The same help page notes that a rate can exceed 100% when several conversions are counted for one interaction.

The catch is the denominator, the number you divide by. You can divide by sessions (visits), by users, by ad clicks, or by visits to one page. Each gives a different rate for the same business. A site with 10,000 sessions from 8,000 users and 300 enquiries has a 3.0% session rate and a 3.75% user rate.

Pick one definition per question and keep it. For a landing page, divide by visits to that page. For an ad, divide by ad clicks. For the whole site, say so. Mixing them is the most common reason two reports disagree.

What counts as a conversion

A conversion is whatever action you decide is valuable: a purchase, a submitted quote form, a booked call or a click on the phone number. In Google Analytics 4 you mark an event as a key event so that it is counted this way, as described in Google’s Analytics Help.

Separate main conversions from micro-conversions. A purchase or a booked call is a main conversion. Viewing pricing or starting a form is a micro-conversion, a step that signals interest but earns no money on its own. Optimise for the main one and use micro-conversions to find where people stall.

A rate for the wrong action misleads. A page that gets many newsletter sign-ups but no sales may have a high rate and no value. Track the action that matches the business goal.

Why small samples mislead

A rate from few visitors is a rough guess, and a small sample size is the cause. Suppose 20 visitors arrive and 2 convert. The rate is 10%, yet a standard 95% confidence interval, calculated with the Wilson method, runs from about 2.8% to 30%. The real rate could be a third of that or three times it.

More data narrows the range. With 200 visitors and 10 conversions (5%), the range is about 2.7% to 9.0%. With 2,000 visitors and 100 conversions (also 5%), it is about 4.1% to 6.0%. The rate is the same in all three cases, but your confidence in it is not.

This is why a week of numbers on a low-traffic site rarely proves anything. It is also why one good Monday is not a trend. Wait until the sample is large enough, or say plainly that the number is uncertain.

Statistical significance, explained plainly

Statistical significance answers one question: how likely is it that a difference this big appeared by chance, if the change did nothing? A common cut-off is 5%. A result below that is called significant. It does not say the change is large or valuable, only that chance is an unlikely explanation.

Worked example. Page A converts 40 of 1,000 visitors (4.0%). Page B converts 52 of 1,000 (5.2%). That looks like a 30% lift. A standard two-proportion test gives a score of about 1.28, well short of the 1.96 needed at the 5% level. The gap is not significant. With these numbers, chance explains it comfortably.

Peeking makes this worse in any A/B test. Evan Miller, in his 2010 article “How Not To Run an A/B Test”, shows that repeatedly checking a running test and stopping when it looks significant inflates false positives. In his example with no real effect, stopping at the first significant reading produced a 26.1% false positive rate. Decide the sample size first and wait.

Minimum detectable effect and how much traffic you need

The minimum detectable effect (MDE) is the smallest change you want a test to be able to catch. It is a business choice: how small an improvement would still be worth the effort? The smaller the MDE, the more visitors the test needs.

Evan Miller’s sample size calculator takes four inputs: baseline conversion rate, MDE, statistical power (the chance of catching a real effect of that size) and significance level. Using the common settings of 80% power and a 5% significance level, a standard formula gives these approximate numbers per variation. Different calculators use different assumptions, so figures vary.

• Baseline 5%, target 6% (a 20% relative lift): about 8,150 visitors per variation.
• Baseline 5%, target 7.5% (a 50% relative lift): about 1,470 per variation.
• Baseline 2%, target 2.5%: about 13,800 per variation.
• Baseline 10%, target 12%: about 3,840 per variation.

What the sample size table tells a business owner

Take the 5% baseline and a 20% relative MDE. To test whether a change lifts it to 6%, you need about 16,300 visitors in total, split across two versions. A page with 2,000 visits a month needs roughly eight months. That test is not practical.

You have three honest options. Aim for a bigger change that has a larger MDE and needs less traffic. Test on a higher-traffic page. Or skip formal testing and use research, as covered in CRO research methods. All three beat a test that never finishes.

Also run a test through whole weeks. Weekday and weekend behaviour differ, so stopping on a Wednesday can bias the result.

Revenue per visitor: the number that decides

Conversion rate ignores how much each conversion is worth. Revenue per visitor (RPV) is total revenue divided by total visitors, and it captures both rate and order value. Suppose version A gets 10,000 visitors, 300 orders at an average of 80, so revenue is 24,000 and RPV is 2.40.

Version B converts better at 3.3% (330 orders) but its average order is 70 because it pushes a cheaper bundle. Revenue is 23,100 and RPV is 2.31. B has the higher conversion rate and makes less money. Without RPV you would ship the wrong version.

Lead-gen businesses can do the same with a simple chain. If 2% of visitors enquire, 25% of enquiries become clients, and a client is worth 4,000, then each visitor is worth 2% × 25% × 4,000 = 20. That estimate is only as good as your close rate and value, so use your own numbers.

Benchmarks: use them with care

Benchmarks vary widely. Unbounce’s 2024 Conversion Benchmark Report, a vendor study of over 41,000 landing pages, gives a median of 6.6% across industries and industry medians from 3.8% to 12.3%. Baymard Institute’s September 2025 aggregation of 50 studies puts average documented cart abandonment at 70.22%.

These are different measures from different samples, so they do not tell you what your rate should be. Price, traffic source, intent and how you define a conversion all move the number. Use benchmarks to check you are in a plausible range, then compare yourself with yourself.