UserProof

UserProof / Free tools / How many visitors to test a change

How many visitors do I need to test a website change? A sample size calculator for founders

Free tool · runs in your browser · Updated November 15, 2026
TL;DRA sample size calculator for a change to your website. From a 3% conversion rate, seeing a 20% change (3.0% to 3.6%) takes 12,675 visits before the change and 12,675 after, with 95% confidence and 80% power, the same formula UserProof's fix check uses. Big changes need far fewer: the Sales.co fix, a bounce rate falling from 76% to 53%, needed 53 visits a side and had 118 before and 85 after. The numbers you type stay in your browser.

For founders who build with a coding agent and ship changes to a live site: from a 3% conversion rate, a 20% change takes 12,675 visits before and 12,675 after; from 5% it takes 7,448. At 200 visits a day that is 64 days after the change. If your page gets less, test bigger changes: a 50% change from 5% needs 1,192 a side.

By Ryan Doyle, founder of UserProof, whose fix check uses this formula to say how much traffic a change would need.

Your numbers

The calculator needs JavaScript and its script hasn't run, so editing changes nothing here. The result below is the example, worked out in full.

Example result

12,675 visits before the change and 12,675 after. That is what it takes to see 3.0% move to 3.6% (a 20% change, 0.6 points) with 95% confidence and 80% power, the standard UserProof's fix check uses.

At 200 visits a day, the after side takes 64 days, longer than the fix check's 14-day window. In 14 days (2,800 visits a side) the smallest change you could see is 1.3 points, 3.0% to 4.3% (43% of today's rate).

Visits and days for the change you entered
How you testVisits neededDays at 200 a day
Before and after (ship, then compare)12,675 before + 12,675 after64 after the change, if you already have 64 days before
A/B split, 50/5012,675 per version, 25,350 in all127

Visits needed at common rates

Visits on each side (before and after, or each version of an A/B split) to see a change of 10%, 20% or 50% of today's rate, at 95% confidence and 80% power.

Visits per side by conversion rate and the change you want to see
Rate today10% change20% change50% change
1%155,23238,8086,210
2%76,83219,2083,074
5%29,7927,4481,192
10%14,1123,528565

A page converting at 2% needs 19,208 visits a side to show a 20% lift, and 76,832 to show a 10% lift. Halve the change and the visits roughly quadruple, because the formula divides by the change squared. Low rates cost more: the same 20% change needs 38,808 visits a side at 1% and 3,528 at 10%.

What your traffic can show in 14 days

UserProof's fix check reads 14 days before a change and up to 14 days after. Turned around, the formula says the smallest change a page can show in that window at a 3% rate:

Smallest change visible in 14 days a side, from a 3% rate
Visits a dayVisits a sideSmallest changeRate moveAs % of today's rate
507002.55 points3.00% to 5.55%85%
2002,8001.28 points3.00% to 4.28%43%
1,00014,0000.57 points3.00% to 3.57%19%
5,00070,0000.26 points3.00% to 3.26%9%

At 50 visits a day only a change of 85% of the rate stands out in two weeks; at 5,000 a day, 9% does. A founder with a small site gets clear reads on big moves (a broken button fixed, a sign-up button moved onto the first screen) and inconclusive reads on small ones (a word in a headline). To find problems rather than measure a rate, five people in a usability test are enough, with no statistics needed.

Worked example and a real one

The example: 3% of visitors sign up today and you want to see a 20% change, so the target is 3.6%, a change of 0.6 points. The formula gives 2 × (1.96 + 0.84)² × 0.03 × 0.97 ÷ 0.006² = 12,675 visits on each side. At 200 a day the after side takes 64 days, and a 50/50 A/B split of the same traffic takes 127 days for both versions to reach it.

The real one: on Sales.co, 90 of 118 visits to the home page bounced before a fix (76%) and 45 of 85 after (53%), a change of 23 points (the fix check page). From 76%, a 23-point change needs 53 visits a side. Sales.co had 118 before and 85 after, so the fix check could call it, with a 95% interval of 10 to 36 points on the drop. The change was large, so a small site could see it in six days.

How it works

The fix check itself picks δ as 20% of the before rate (and never under 0.2 points) when it reports the traffic it would need. Here you choose δ. The code is the open file sample-size-calc.js.

Limits: what it can't tell you

Then check the change on real visits

Once a change has its visits, the fix check calculator reads the result from your before and after counts. UserProof does both on recorded visits. When your coding agent ships a fix, it compares the page before and after against your other pages, says when there is too little traffic, and tells the agent whether the fix worked (how the fix check works). Free for 3 sites and 5,000 recorded visits a month; Pro is $29 a month for 50,000; Scale is $99 for 250,000 (pricing).

To put a dollar figure on the change you are sizing, the conversion rate increase calculator turns a lift into money a month and a year.

Related: the funnel calculator, the dead click finder, all free tools.

Sources, read November 15, 2026

Cite this

UserProof, "How many visitors do I need to test a website change?", https://getuserproof.com/how-many-visitors-to-test-a-website-change, updated November 15, 2026. Visits per side = 2 × (1.96 + 0.84)² × p(1−p) ÷ δ², at least 30; from 3% a 20% change needs 12,675 a side.

FAQ

How many visitors do I need to test a website change?

It depends on your rate today and the smallest change you care about. From a 3% conversion rate, seeing a 20% change (to 3.6%) takes 12,675 visits on each side. From 5% it takes 7,448, and from 10% it takes 3,528. A 50% change from 5% takes 1,192. Halving the change you want to see roughly quadruples the visits.

Is 100 visitors enough to test a change?

Only for a very large change on a high rate. 100 visits a side can show a change of about 20 points from a 50% rate, or a bounce rate falling from 76% to 53%, which needed 53 a side. For a 3% signup rate, 100 visits a side can't tell a 20% lift from noise: that takes 12,675.

How long should I run a before and after test?

Divide the visits needed by the visits the page gets a day. At 200 a day, a 20% change from 3% needs 64 days after the change, and the same number of days before. UserProof's fix check reads 14 days before and up to 14 days after, so a change that needs longer reads as inconclusive there until traffic catches up.

Does a before and after test need fewer visitors than an A/B test?

Per side it is the same number. An A/B split puts half your traffic on each version, so in calendar days it takes twice as long as one after period. A before and after read reuses the visits you already had before the change, but a holiday or a second change in the same weeks lands in the result, which an A/B split avoids.

What confidence and power does the calculator use?

95% confidence, two-sided, and 80% power: the 1.96 and 0.84 in the formula. Those are the values in UserProof's fix check code. Power is the chance a real change of the size you entered shows up as a clear result.

Does the calculator store or send my numbers?

No. It runs in your browser, makes no requests and saves nothing. The page has the same visit analytics as the rest of this site, and the form and the result are marked so the recorder skips them.