Price test sample size calculator

How many visitors does a price test need? Sample size for revenue per visitor, how many days at your traffic, and how an adaptive test compares.

Free, no signup, runs in your browser. Built by Leverprice, which tests prices on real visitors for Stripe, Polar, Lemon Squeezy and Dodo Payments.

Count pricing-page or checkout visitors, not all site traffic.

10% means the test price earns 10% more per visitor.

Prices in the test
Prices in the test

A fixed even-split test needs

95,564

visitors per price, 191,129 in total. At 500 visitors a day that is about 13 months (95% confidence, 80% power).

1.64%

Conversion needed at $39

$0.58

Revenue per visitor now

$0.64

At the test price

Short answer

A price test needs enough visitors to tell revenue per visitor apart, which takes far more traffic than a conversion test because most visitors spend nothing. Enter your traffic, conversion rate and test price to get visitors per price, days to finish, and our simulated estimate for an adaptive test.

The formula

For each price, revenue per visitor is price × conversion and its variance is price² × conversion × (1 − conversion). The visitors needed per price for a two-sided test at 95% confidence and 80% power are:

n = (1.96 + 0.84)² × (σ²₁ + σ²₂) / (μ₂ − μ₁)²

With $29 at 2% conversion and a test price that earns 10% more per visitor, each price needs roughly 110,000 visitors when the test price is $49, and fewer when the two prices are closer, because a closer price has lower variance. Change the inputs to see how strongly the difference you want to detect drives the result.

Fixed tests and adaptive tests answer different questions

A fixed test is designed to measure the lift precisely, which is why it is large. An adaptive engine, like Leverprice, moves traffic toward the leader as it learns and stops when the expected loss of picking the leader is tiny. That is a different stopping rule, so its numbers are not directly comparable. We ran 960 simulations of our production engine to show how long that takes at different volumes: How long should a price test run?

Frequently asked questions

Why does a price test need so many more visitors than a normal A/B test?

A price test compares revenue per visitor, not clicks. Most visitors spend nothing and a few spend the full price, so the number is very noisy, and a 10% difference in its average takes far more data to see than a 10% difference in a click rate of 30%.

What does 95% confidence and 80% power mean here?

If the prices really earn the same, a test of this size declares a difference only 5% of the time. If the test price really earns as much more as you entered, the test detects it 80% of the time.

Can I test with fewer visitors?

Yes: test prices that are further apart, test the plan that brings in most of your revenue, or use an adaptive test that stops once the remaining difference is too small to matter. Our simulation estimate on this page shows what that looked like for the Leverprice engine.