Glossary
Conversion rate: definition and formula
What the number means, how to calculate it without fooling yourself, and how it behaves inside an A/B test.
Definition: Conversion rate is the share of visitors who complete a chosen action, such as a purchase or a sign-up. Divide conversions by visitors over the same period and multiply by 100: 300 conversions from 10,000 visitors is a 3% conversion rate. In an A/B test each variant has its own rate and the test compares them.
The conversion rate formula
conversion rate = conversions ÷ visitors × 100
Both numbers must cover the same period and the same page or funnel. Conversions are usually counted once per visitor, so a customer who buys twice in the period is one conversion, not two.
How to calculate it, step by step
- Define the conversion. One specific action: an order placed, a form submitted, a trial started.
- Pick the period and the scope. A calendar month for one landing page, say.
- Count the visitors to that scope in that period. Unique visitors, not pageviews.
- Count the conversions from those visitors, once each.
- Divide and multiply by 100.
The conversion rate calculator does the arithmetic and adds what a given uplift would be worth in conversions and revenue.
Visitors, sessions or pageviews?
The denominator changes the number. 300 conversions from 10,000 visitors is 3%; the same 300 from those visitors' 14,000 sessions is 2.1%; from 40,000 pageviews it is 0.75%. None is wrong, but they cannot be compared with each other. Pick one and keep it. A/B testing tools count visitors, because assignment is per visitor.
What is a good conversion rate?
Published benchmarks vary by industry, by traffic source and, above all, by what each site counts as a conversion, so a figure from a report tells you little about your own page. The comparisons that matter are your own rate over time and, in a test, the rate of one variant against another measured on the same visitors in the same weeks.
Conversion rate in an A/B test
Each variant gets a conversion rate from its own visitors. The difference between them is the uplift, and because both rates are measured on samples, the question is whether the gap is bigger than chance. That is what statistical significance answers.
A worked example
A checkout page had 4,000 visitors and 120 orders last month: 120 ÷ 4,000 × 100 = 3.0%. A variant tested against it converted 150 of 4,000: 3.75%. The relative uplift is (3.75 - 3.0) ÷ 3.0 = 25%, the absolute uplift is 0.75 percentage points. Whether that difference is real is a separate question; at these sample sizes it is not yet significant.
Common mistakes
- Mixing denominators: visitors one month, sessions the next.
- Counting a conversion every time the thank-you page loads, including refreshes and repeat buyers.
- Comparing a rate from a campaign week with a rate from a quiet week and calling the difference an improvement.
- Chasing conversion rate while revenue per visitor falls, because the variant attracts smaller orders.
In AB Test WP
AB Test WP counts one impression and one conversion per visitor per test, so repeat clicks do not inflate a variant's rate, and it shows the rate, visitors and conversions for every variant side by side.
Related: Conversion rate calculator · What to A/B test first · all glossary terms.
Common questions
Is conversion rate calculated on visitors or sessions?
Either works if you are consistent, but visitors is the usual base for A/B testing because each visitor is assigned to one variant. State which one you used whenever you report a rate.
What counts as a conversion?
Whatever action the page exists to produce. It should be specific, measurable and close to the money: an order is better than a click on "Add to cart", which is better than a scroll.
How much does conversion rate vary on its own?
More than people expect. A page at 3% can show 2.4% one week and 3.6% the next with no change at all, especially on small numbers. That random variation is the reason tests need a planned sample size.
Run the numbers on your own test
Free calculators, no email wall: significance, sample size, conversion rate and SRM.