TapCub is live — analytics, insights and live chat, free on one platform

Free tool Conversion rate calculator and A/B test significance

Visitors and conversions in, rate out — with a confidence interval. Add a second group to compare variants and see whether the difference is more than noise.

  • Rate and 95% confidence interval
  • Two-proportion z-test for A/B
  • Runs locally

Enter visitors and conversions

Use the same definition for both groups: the same visitor count (people, not page views) and the same converting event.

Calculator

Group A

Control

Conversion rate

21.2%

±1.6 pp · 95% confidence interval

Group B

Variant · optional

Conversion rate

24.0%

±1.7 pp · 95% confidence interval

A vs B

Two-proportion z-test with pooled variance, two-tailed. Significant at p < 0.05.

Absolute difference+2.8 pp
Relative lift+13.3%
z-score2.34
p-value0.019
B converts better than A, and the difference is statistically significant.

The test assumes each visitor is counted once and the two groups were split at random. It says nothing about whether the lift matters to the business.

Runs entirely in your browser. Nothing you type here is uploaded or stored by TapCub.

Share these inputs: the values go after the # in the link, so they never reach a server.

Version v2.0Updated Update history

NewRebuilt on the new tools template, with links to the sample size, p-value, test duration and SRM tools.

Read the result

What each number is telling you

With one group you get a rate and its interval. Add a second and the comparison panel answers "is the difference real?".

The conversion rate

The big number is conversions ÷ visitors for that group. Read it next to the interval below it, never alone — a 21% rate from 20 visitors and from 20,000 are not the same claim.

The ± confidence interval

The true rate is very likely within this band. Overlapping bands between A and B are a hint the difference may be noise; the z-test is the actual decision.

Absolute vs relative lift

Absolute is in percentage points (21% → 24% is +3 pp). Relative restates it against the baseline (+14%). Headlines quote the relative number; budgets should use the absolute one.

The verdict line

Green means the z-test cleared p < 0.05: a difference this large is unlikely to be chance. Amber means not yet — keep collecting, or accept you cannot tell these variants apart.

Statistical significance is not business significance. A real but tiny lift can still fail to pay for the change that produced it.

How to use it

Three decisions before you type a number

The arithmetic is instant. Getting the inputs right is the work.

  1. 1

    Decide what counts as a visitor

    Unique people who had the chance to convert — not page views, and ideally with bots removed. 2,400 product viewers in the sample, not 48,213 site visitors.

  2. 2

    Decide what counts as a conversion

    One event, counted once per visitor: a paid order, a submitted form, a completed sign-up. The sample uses 508 paid orders.

  3. 3

    Keep the time window identical

    Both groups should cover the same days. A variant that ran over a weekend is not comparable to a control that ran on weekdays.

Formula

The rate, the interval and the test

The rate alone hides how many people it is based on. The confidence interval puts the sample size back in, and the z-test asks whether two intervals really disagree.

  • Rate = conversions ÷ visitors
  • Interval width shrinks with the square root of the sample size — four times the traffic halves the width
  • The z-test pools both groups to estimate variance under "no difference"
  • p < 0.05 means a difference this large would appear by chance less than 1 time in 20

Conversion rate

p = conversions ÷ visitors

Sample: 508 ÷ 2,400 = 21.2%

95% confidence interval (Wald)

p ± 1.96 × √( p (1 − p) ÷ n )

Sample: 21.2% ± 1.6 pp → 19.5% to 22.8%

Two-proportion z-test

z = (p₂ − p₁) ÷ √( p̄ (1 − p̄) (1/n₁ + 1/n₂) ), p̄ = (x₁ + x₂) ÷ (n₁ + n₂)

Sample: z = 2.34, p = 0.020 → significant at 5%

Worked example

How to calculate conversion rate and significance by hand

The default data, step by step: group A had 508 conversions from 2,400 visitors, group B 571 from 2,380. The last column is what the calculator shows above.

Conversion rate formula and two-proportion z-test on the default data
StepFormulaResult
Conversion rate A508 ÷ 2,400 × 100%21.2%
Conversion rate B571 ÷ 2,380 × 100%24.0%
Absolute lift23.99% − 21.17%+2.8 pp
Relative lift(23.99% − 21.17%) ÷ 21.17%+13.3%
Pooled rate p̄(508 + 571) ÷ (2,400 + 2,380)22.57%
Standard error√( 0.2257 × 0.7743 × (1/2,400 + 1/2,380) )0.01209
z and p-valuez = 0.02825 ÷ 0.01209, two-tailed pz = 2.34, p = 0.020

Keep full precision until the last step: rounding the rates first can shift z in the second decimal. p = 0.020 is below 0.05, so B’s lift is significant at the 5% level.

Common mistakes

How conversion rates go wrong

Most disagreements about a rate are disagreements about the denominator.

Page views as the denominator

One person reloading five times becomes five chances to convert. Count visitors, not views.

Bots left in

Crawlers never convert. Leaving them in the denominator drags the rate down and makes campaigns look worse than they are.

Stopping the test early

Peeking daily and stopping at the first p < 0.05 inflates false positives. Fix the sample size or duration first.

Different windows

A rate from a sale week against a rate from a normal week measures the sale, not the change you made.

FAQ

Conversion rate questions

Still have a question?

Chat with the team behind TapCub — we usually reply within a few hours on business days.

How do you calculate conversion rate?

Conversion rate = conversions ÷ visitors × 100%. 508 orders from 2,400 visitors is 508 ÷ 2,400 = 21.2%. Use people (or sessions, consistently) as the denominator rather than page views, and count each converting person once. The same formula gives an add-to-cart rate (carts ÷ visitors) or a sign-up rate.

How do I know if an A/B test result is statistically significant?

Enter visitors and conversions for both groups. The calculator runs a two-proportion z-test (pooled variance, two-tailed): z = (p₂ − p₁) ÷ √(p̄(1 − p̄)(1/n₁ + 1/n₂)). If p is below 0.05, the difference is statistically significant at the 5% level. On the default data z = 2.34 and p = 0.020, so B wins.

How many visitors do I need?

It depends on the baseline rate and the lift you want to detect. Detecting a 10% relative lift on a 20% baseline (20% → 22%) needs about 6,510 visitors per group at 5% significance and 80% power; smaller lifts need far more. The A/B test sample size calculator works it out for your numbers.

What does "not significant" mean?

That the observed difference could plausibly be chance. It does not mean the variants are equal — only that you have not shown they differ.

Why is my rate different from my analytics tool?

Usually the denominator: sessions vs visitors, bots included or not, and whether the converting event is counted once per person. Match the definitions before comparing.

Can I stop the test as soon as it turns significant?

No. The z-test assumes you look once, at a sample size decided in advance. Checking every day and stopping at the first p < 0.05 makes false positives far more likely than 5%. Pick the sample size first, run the test to that size, then read the result.

Changelog

Update history

New features, fixes and data refreshes for this tool, newest first.

  1. · v2.0

    ImprovedRebuilt on the new tools template, with links to the sample size, p-value, test duration and SRM tools.

  2. · v1.0

    LaunchedFirst release.

Missing something or found an error? Tell us — requests that come up often go into the next update. All tool updates

Next step

See the rate for every step, automatically

Connect your site and TapCub builds the funnel from the events it already captures. Compare segments, channels and time ranges without a calculator.

See your data clearly. Find your growth.

Every click, backed by data. Install one line of code and see your first numbers in a minute.

No credit card · Free plans for analytics and chat · Cookieless analytics