Conversion Rate Calculator
Conversion rate, A/B test significance, sample size and traffic goals
Conversion rate = conversions ÷ visitors × 100. 285 orders from 12,500 sessions is a 2.28% conversion rate, one order for every 43.9 visits. To judge an A/B test, compare the two rates with a two-proportion z-test: 3.00% against 3.60% on 14,000 visitors each gives p = 0.0050, significant at 95% confidence.
Last updated October 2026
Method: Conversion rate = conversions ÷ visitors. The 95% range is the Wilson score interval. A/B results use the pooled two-proportion z-test from the NIST/SEMATECH e-Handbook of Statistical Methods. Sample sizes use the normal-approximation formula for two independent proportions (Fleiss, Tytun and Ury, without continuity correction).
Included: Conversion rate, visitors per conversion, 95% range, revenue per visitor, average order value, cost per conversion, relative and absolute lift, z-score, p-value, interval for the difference, visitors per variant at 90/95/99% confidence and 80/90% power, test length, and visitors and budget for a sales goal.
Not included: Bayesian and sequential testing, corrections for peeking or for more than two variants, Fisher's exact test for tiny samples, and significance tests on revenue per visitor.
🎯 What do you want to calculate?
🎯 Conversion rate
📈 What a higher conversion rate is worth on the same traffic
| Relative lift | New rate | Extra conversions | Extra revenue |
|---|---|---|---|
| +10% | 2.51% | 28.5 | $2,422.50 |
| +25% | 2.85% | 71.3 | $6,056.25 |
| +50% | 3.42% | 142.5 | $12,112.50 |
Extra revenue assumes the average order value stays at $85.00. The 95% range is a Wilson score interval: with 12,500 visitors, rates anywhere in that band are consistent with what you observed.
Conversion rate calculator: formula, A/B testing and sample size
A conversion rate calculator divides the visitors who took an action (an order, a sign-up, a lead form) by all visitors who had the chance. 285 orders from 12,500 sessions is a 2.28% conversion rate. This tool goes three steps further: it puts a 95% range around the rate, tests whether variant B really beats variant A, and tells you how much traffic a test or a sales goal needs before you pay for it.
For the money side of the same campaign, the ROAS Calculator compares revenue with ad spend, and the Contribution Margin Calculator shows whether each converted sale actually earns money.
The conversion rate formula
Every version of the metric uses the same fraction; only the definitions of the top and bottom change.
Conversion rate = conversions ÷ visitors × 100 Revenue per visitor = revenue ÷ visitors Average order value = revenue ÷ conversions Revenue per visitor = conversion rate × average order value The last line is an identity rather than a new formula, and it is the most useful one for decisions. A store earns more per visitor by converting more of them or by selling more to each buyer, so a change that lowers the conversion rate is not automatically bad if the average order rises by more.
Worked example 1: an online store's monthly rate
In March a store recorded 12,500 sessions, 285 orders and $24,225 in revenue.
- Conversion rate: 285 ÷ 12,500 × 100 = 2.28%, one order per 43.9 sessions.
- Average order value: $24,225 ÷ 285 = $85.00.
- Revenue per visitor: $24,225 ÷ 12,500 = $1.94, the same as 2.28% × $85.00.
- 95% range for the true rate: 2.03% to 2.56%.
March's 2.28% is one draw from the store's underlying rate. With 12,500 sessions, anything from about 2.0% to 2.6% would have been unsurprising, so an April reading of 2.45% is not proof of progress on its own. On the same traffic, a 10% higher rate (2.51%) would add 28.5 orders and $2,422.50 a month.
Worked example 2: a landing page fed by paid clicks
A lead-generation page received 2,400 ad clicks in two weeks and collected 186 completed forms. The clicks cost $3,000.
- Conversion rate: 186 ÷ 2,400 × 100 = 7.75%, one lead per 12.9 clicks.
- Cost per click: $3,000 ÷ 2,400 = $1.25.
- Cost per lead: $3,000 ÷ 186 = $16.13.
- 95% range: 6.75% to 8.89%.
A focused landing page can convert at a higher rate than a whole store because the page has one job and the visitors arrived with intent, so compare it with its own history rather than with your store average. Count clicks that actually loaded the page rather than clicks the ad platform reports, or the rate will look worse than it is.
What counts as a visitor, a conversion and a good rate
Most disagreements about conversion rates are disagreements about definitions. Sessions count visits, so a shopper who returns three times counts three times; users count people, as far as cookies allow; clicks count ad interactions, some of which never load the page. Orders, sign-ups and demo requests are all conversions, but each page should have one headline goal.
There is no official yardstick for a good rate. The Census Bureau measures how much Americans buy online (e-commerce was 17.1% of total retail sales in the second quarter of 2026, in the seasonally adjusted estimate released August 18, 2026), but no government source publishes conversion benchmarks. Benchmark against yourself: the same pages, season and channels, with revenue per visitor beside the rate.
How precise is a conversion rate?
The calculator puts a 95% range around every rate with the Wilson score interval. The NIST/SEMATECH e-Handbook presents it in place of the simple normal-approximation interval, which can produce impossible values below zero. For an observed 3% rate:
| Visitors | Conversions | 95% range for the true rate |
|---|---|---|
| 100 | 3 | 1.03% to 8.45% |
| 500 | 15 | 1.83% to 4.89% |
| 1,000 | 30 | 2.11% to 4.25% |
| 2,000 | 60 | 2.34% to 3.84% |
| 5,000 | 150 | 2.56% to 3.51% |
| 10,000 | 300 | 2.68% to 3.35% |
| 50,000 | 1,500 | 2.85% to 3.15% |
| 100,000 | 3,000 | 2.90% to 3.11% |
At 100 visitors a 3% rate is consistent with anything from about 1% to 8.5%, which is useless for decisions. The range narrows with the square root of the sample, so halving its width takes four times the traffic. The same arithmetic makes A/B tests expensive.
A/B test significance with a two-proportion z-test
To compare two variants, the calculator runs the pooled two-proportion z-test described in the NIST/SEMATECH e-Handbook:
Pooled rate p̂ = (xA + xB) ÷ (nA + nB) SE = √[ p̂ × (1 − p̂) × (1/nA + 1/nB) ] z = (pB − pA) ÷ SE A checkout test sent 14,000 visitors to each arm. A converted 420 (3.00%), B converted 504 (3.60%). The pooled rate is 3.30%, the standard error is 0.2135 percentage points, and z = 0.60 ÷ 0.2135 = 2.81. The two-sided p-value is 0.0050, so the 20% relative lift is significant at 95% and even at 99%. The 95% interval for the true difference runs from +0.18 to +1.02 points.
Shrink the test to 1,200 visitors per arm with 36 and 45 conversions. The observed lift is bigger (+25%), yet z = 1.02 and p = 0.309, a gap small samples produce by chance all the time. The interval, −0.69 to +2.19 points, includes zero, and the calculator notes that a lift of that size needs about 9,100 visitors per variant to detect reliably.
How many visitors an A/B test needs
Sample size comes from four inputs fixed before the test: the baseline rate p1, the target rate p2 implied by your minimum detectable effect, the confidence level and the power (the chance of catching a true effect of that size). The formula is the two-sample version of the NIST sample-size formula for proportions:
n = [ z1−α/2 √(2p̄(1 − p̄)) + z1−β √(p1(1 − p1) + p2(1 − p2)) ]² ÷ (p2 − p1)² p̄ = (p1 + p2) ÷ 2 For a 3.00% baseline and a 20% minimum lift (3.60%) at 95% confidence and 80% power, the z values are 1.95996 and 0.84162 and p̄ = 3.30%. Then n = (0.49515 + 0.21259)² ÷ 0.006² = 13,913.6, rounded up to 13,914 visitors per variant, 27,828 in total, or 14 days at 2,000 visitors a day. At 90% power the requirement is 18,626; at 99% confidence, 20,704. Tools that use unpooled variance or a continuity correction give slightly different numbers.
The table applies the formula at 95% confidence (two-sided) and 80% power. Lifts are relative: +10% on a 2% baseline means 2.0% to 2.2%.
| Baseline | +5% | +10% | +15% | +20% | +30% | +50% |
|---|---|---|---|---|---|---|
| 1% | 637,010 | 163,095 | 74,193 | 42,693 | 19,827 | 7,750 |
| 2% | 315,206 | 80,682 | 36,693 | 21,109 | 9,798 | 3,826 |
| 3% | 207,938 | 53,211 | 24,193 | 13,914 | 6,455 | 2,518 |
| 5% | 122,124 | 31,234 | 14,193 | 8,158 | 3,780 | 1,471 |
| 10% | 57,763 | 14,751 | 6,693 | 3,841 | 1,774 | 686 |
| 20% | 25,583 | 6,510 | 2,943 | 1,683 | 772 | 294 |
Low baselines are expensive: a 10% lift costs 163,095 visitors per variant at 1% but 14,751 at 10%. Small lifts are worse: on a 3% baseline, a 5% lift needs 207,938 per variant, about 15 times the traffic for a 20% lift.
Visitors needed for a sales goal
Planning works backward from the sales you need: visitors needed = target conversions ÷ conversion rate, rounded up. To win 250 sales at 2.5% you need 10,000 visitors. At $1.20 a click that is a $12,000 budget, or $48.00 per sale, against $21,250 of revenue at an $85 average order. Lift the rate to 3.125% and the same 250 sales need only 8,000 visitors and $9,600.
| Conversion rate | 50 sales | 100 sales | 500 sales | 1,000 sales |
|---|---|---|---|---|
| 0.5% | 10,000 | 20,000 | 100,000 | 200,000 |
| 1% | 5,000 | 10,000 | 50,000 | 100,000 |
| 2% | 2,500 | 5,000 | 25,000 | 50,000 |
| 3% | 1,667 | 3,334 | 16,667 | 33,334 |
| 5% | 1,000 | 2,000 | 10,000 | 20,000 |
| 10% | 500 | 1,000 | 5,000 | 10,000 |
How to use this calculator
- Conversion rate: enter visitors and conversions for the same period. Add revenue for revenue per visitor and average order value, and traffic cost for the cost per conversion. The table underneath shows what a 10%, 25% or 50% lift would add.
- A/B test: enter visitors and conversions for each variant, then pick 90%, 95% or 99% confidence and a two-sided or one-sided hypothesis. Read the verdict, p-value and interval.
- Sample size: enter your baseline rate and the smallest lift worth detecting, as a relative percentage or in points, plus daily traffic for the test length.
- Traffic goal: enter the conversions you want and the rate you expect, with cost per visitor and order value for budget and revenue.
The tool is built for store owners checking monthly performance, growth teams planning tests, marketers budgeting paid campaigns, founders sizing a launch, and students working on proportion tests.
Key terms
- Relative lift: (pB − pA) ÷ pA. Going from 3.00% to 3.60% is +20%.
- Percentage points: the plain difference pB − pA. The same move is +0.60 points.
- p-value: the probability of a gap at least this large if both variants truly convert at the same rate.
- Confidence level: 1 − α. At 95%, a result counts as significant when p is below 0.05.
- Power: the probability of detecting a true effect of the chosen size; 80% is a common floor.
- Minimum detectable effect: the smallest true lift the test is sized to catch.
- Sample ratio mismatch: an observed traffic split that differs from the configured one, usually a tracking or assignment bug.
What moves the numbers, and tips for cleaner tests
- Traffic mix: paid, email and returning visitors convert differently, so a shift in channels moves the site-wide rate without any change to the site.
- Season and promotions: compare like periods with like periods.
- Speed, errors and price: slow pages, checkout bugs and shipping costs often matter more than button colors.
- Plan first: fix the sample size, the primary metric and the decision rule before launch, and keep a log of every test.
Limitations
The z-test and the sample-size formula rely on the normal approximation, which NIST describes as suitable for reasonably large samples; with a handful of conversions per group, Fisher's exact test is the better tool. The calculator assumes each visitor is counted once and assigned at random, and it does not adjust for peeking, extra variants or several metrics tested at once. It treats every conversion as equal, so it cannot test revenue per visitor, which needs a test on means. The Wilson range covers sampling noise only, not tracking errors or bots.
Related calculators
- Price your impressions with the CPM Calculator and judge ad returns with the ROAS Calculator.
- Express a month-over-month change in the rate with the Percentage Change Calculator.
- Look up z values for other confidence levels with the Critical Value Calculator, or standardize a result with the Z-Score Calculator.
- See how many sales cover your fixed costs with the Break-Even Calculator, and how long a marketing investment takes to earn back with the Payback Period Calculator.
Sources
- NIST/SEMATECH e-Handbook of Statistical Methods, section 7.3.3: How can we determine whether two processes produce the same proportion of defectives? (pooled two-proportion z-test, normal approximation for reasonably large samples, Fisher's exact test for small ones).
- NIST/SEMATECH e-Handbook, section 7.2.4.1: Confidence intervals for a proportion (Wilson score interval).
- NIST/SEMATECH e-Handbook, section 7.2.4.2: Sample sizes required (normal-approximation sample size for proportions).
- Fleiss JL, Tytun A, Ury HK. A simple approximation for calculating sample sizes for comparing independent proportions. Biometrics. 1980;36:343-346.
- U.S. Census Bureau: Quarterly Retail E-Commerce Sales, 2nd quarter 2026 (released August 18, 2026).
⚠️ Common mistakes & edge cases
Mixing sessions and users
285 orders over 12,500 sessions is 2.28%, but if those sessions came from 9,000 users the per-user rate is 3.17%. Both are valid and neither is comparable with the other. Keep one denominator across reports and tests.
Stopping the moment the result looks significant
Checking a running test every day and stopping at the first p below 0.05 turns a 5% false-positive rate into a much larger one. Decide the sample size first and read the result once it is reached.
Reporting a relative lift as points
3.00% to 3.60% is +0.60 percentage points and +20% relative. Calling it a 20-point gain overstates the change more than thirty times. Always say which one you mean.
Declaring a winner from a few dozen conversions
A +25% lift on 1,200 visitors per arm has a p-value of 0.309. With fewer than 5 conversions in a group the normal approximation breaks down altogether; the calculator flags that case.
Testing many variants or metrics at once
Each extra comparison at 95% confidence adds its own 5% chance of a false winner. With five independent comparisons, the chance of at least one false positive is about 23%. Name one primary metric in advance.
Ignoring a lopsided traffic split
If a 50/50 test shows 30% of visitors in A, assignment or tracking is broken, and the comparison is no longer between like groups. Fix the setup before reading any lift.
❓ Frequently asked questions
How do you calculate conversion rate?
Divide conversions by visitors and multiply by 100. An online store with 285 orders from 12,500 sessions has a conversion rate of 285 ÷ 12,500 × 100 = 2.28%, about one order for every 43.9 sessions. Both numbers must cover the same pages and the same dates, or the rate means nothing.
What is a good ecommerce conversion rate?
There is no official benchmark. Government statistics track online sales, not conversion rates, and the averages quoted online come from analytics vendors' own client samples, which differ by industry, price point, device and traffic mix. The useful comparison is your own history: the same store, season and channels. A rate that rises while revenue per visitor also rises is a real improvement.
How do I calculate a landing page conversion rate?
Count the visitors who reached the page and the visitors who completed its goal (a form, a sign-up, a purchase) over the same dates, then divide. A paid landing page with 2,400 clicks and 186 leads converts at 7.75%. If those clicks cost $3,000, each lead cost $3,000 ÷ 186 = $16.13.
What is the difference between conversion rate and revenue per visitor?
Conversion rate counts how many visitors buy; revenue per visitor measures what each visitor is worth. They are linked by average order value: revenue per visitor = conversion rate × average order value. A store converting 2.28% with an $85 average order earns $1.94 per visitor. A price increase can lower the conversion rate and still raise revenue per visitor.
How do I know if my A/B test result is statistically significant?
Compare the two rates with a two-proportion z-test and read the p-value. With 420 orders from 14,000 visitors in A (3.00%) and 504 from 14,000 in B (3.60%), z = 2.81 and the two-sided p-value is 0.0050. That is below 0.05, so the 20% lift is significant at 95% confidence. A p-value above your threshold means the data cannot rule out zero difference.
How many visitors do I need for an A/B test?
It depends on the baseline rate, the smallest lift you care about, the confidence level and the power. To detect a move from 3.00% to 3.60% (a 20% relative lift) at 95% confidence and 80% power you need 13,914 visitors per variant, 27,828 in total. Detecting a 10% lift from the same baseline takes 53,211 per variant.
Should I use a one-sided or a two-sided test?
Use two-sided unless you decided in advance that a worse B would be treated exactly like no change. A one-sided test needs fewer visitors (10,960 instead of 13,914 per variant for 3.00% to 3.60%) because it ignores the possibility that B performs worse, which is the risk most site changes actually carry.
How long should an A/B test run?
Until it reaches the sample size you planned, in whole weeks so every weekday is represented equally. At 2,000 visitors a day, a test that needs 27,828 visitors takes 14 days. Stopping the moment the p-value dips below 0.05 inflates false positives, because a p-value checked every day will cross the line by chance sooner or later.
Why do analytics tools report different conversion rates?
They count different things. One tool divides orders by sessions, another by users, a third by ad clicks. They filter bots and internal traffic differently, assign purchases to different days, and miss visitors who block tracking. Pick one definition, write it down, and compare rates only within it.
💡 Good to know
Significant does not mean large
The checkout test above is significant, yet its interval of +0.18 to +1.02 points means the true relative lift could be as small as 6% or as large as 34%. Plan rollouts around the low end.
Two levers, one result
Revenue per visitor is conversion rate times average order value: 2.28% × $85.00 = $1.94. Raising either by 10% lifts revenue per visitor by the same 10%, so bundles and free-shipping thresholds compete with checkout tweaks for attention.
Conversion gains cut ad budgets
250 sales at 2.5% need 10,000 visitors; at 3.125% they need 8,000. At $1.20 a click, that quarter-higher rate saves $2,400 for the same sales.
Related Calculators
ROAS Calculator
Calculate return on ad spend
CPM Calculator
Calculate cost per thousand impressions
Contribution Margin Calculator
Contribution margin per unit, ratio and break-even
Break-Even Calculator
Find the units you must sell to break even
Percentage Change Calculator
Calculate the percent change and difference between values
Z-Score Calculator
Calculate the z-score of a value