A/B Test Revenue Impact Calculator
Every calculator like this one assumes the lift you type in is real and permanent. This one does not, which is why its per-test number is smaller than the one you will see elsewhere, and why its programme number is bigger.
Your store
Sessions to the store, not to one template. This drives the baseline everything else is a percentage of.
Sitewide, as your analytics reports it. Most ecommerce stores sit between 1.5% and 3%.
Revenue divided by orders, before shipping and tax if you can separate them.
The relative improvement in revenue per visitor a single test is meant to deliver.
The same maths with the probability of winning in it, plus what a year of testing adds up to.
How many experiments actually reach a decision each month, not how many are on the backlog.
The share of tests that produce a winner worth shipping. Ours runs around one in four or five.
What a winner actually moves the metric by. Durable winners are usually low single digits.
The 20% win rate and 4% average winning lift are our own published figures, not an industry average: we have said for years that roughly one test in four or five wins, and the four winners in our public checkout programme were +6.2%, +4.9%, +3.1% and +2.2%. Change them to your own numbers if you have them. Where these numbers come from
Where you are now
- Monthly revenue
- $149,760
- Revenue per visitor
- $1.87
- Monthly orders
- 1,920
One test
6.25x smaller.
A year of testing
- Added over twelve months
- $173,000
- Lift in force by month 12
- 18.4%
- Monthly run rate at year end
- $27,600
This is bigger than the number above because it is a programme rather than one test. The per-test assumption did not get more generous.
| Month | Cumulative |
|---|---|
| 1 | $1,200 |
| 2 | $4,790 |
| 3 | $10,800 |
| 4 | $19,200 |
| 5 | $30,000 |
| 6 | $43,100 |
| 7 | $58,700 |
| 8 | $76,700 |
| 9 | $97,000 |
| 10 | $120,000 |
| 11 | $145,000 |
| 12 | $173,000 |
What different lifts are worth
| Lift | Per month | Per year |
|---|---|---|
| 2% | $3,000 | $35,900 |
| 5% | $7,490 | $89,900 |
| 10% | $15,000 | $180,000 |
On 80,000 monthly sessions at 2.4% converting and $78 average order value, a 5% lift is worth $89,900 a year if it lands. Expected value once the 20% win rate is in the number: $14,400 a year per test. A programme of 2 tests a month adds about $173,000 across twelve months. Calculated with the Optyv revenue impact calculator.
Why the per-test number is smaller and the programme number is bigger
Type a 5% lift into any other calculator on this subject and it will multiply your revenue by 5% and then by twelve. That arithmetic is correct and the answer is still wrong, because it quietly assumes something nobody would agree to out loud: that the test wins. Most tests do not. If roughly one in five produces a winner worth shipping, then the expected value of running one test is not the lift, it is the lift multiplied by the chance of getting it.
At our defaults that is the difference between a 5% lift and an expected 0.8%, so the honest per-test figure comes out roughly six times smaller than the one you will be shown elsewhere. That is the entire gap, and it is not pessimism. It is the same number every experienced team already applies in their head when a vendor shows them a projection.
Then the second panel goes the other way, and it is worth being clear about why rather than letting it look like a sleight of hand. A programme running two tests a month for a year is 24 tests, not one. Each winner keeps paying after it ships, so their lifts stack. Twenty-four attempts at a 0.8% expected value, accumulating, comes out larger than one optimistic test held for twelve months. The per-test assumption did not get more generous between the two panels. There are simply more tests in the second one.
This is the honest shape of testing economics, and it is why programmes beat projects. No single test is worth much in expectation. A year of them is worth a great deal, and almost all of that value comes from the compounding rather than from any one result. It also explains why a programme that stops after a bad quarter throws away most of its return: the winners it already shipped keep paying, but nothing new is being added to the stack.
Two more things this calculator does that the others do not. It ramps: a winner found in month nine only pays for the rest of the year, not for all of it, so the accumulation is credited from when the test actually finished. And it caps the annual projection at 30%, because compounding assumptions stop being credible past there, and so do the people making them. If you want the assumptions behind the defaults, they are published in full, including the numbers we decline to publish and why.
The formulas, in fullNothing here is hidden. Every number on the page comes from five inputs and two lines of arithmetic.
- Baseline monthly revenue is sessions x conversion rate x average order value. Revenue per visitor is conversion rate x average order value, which is the metric a lift is applied to.
- A single test at an asserted lift is worth baseline revenue x lift, per month. Held for twelve months, that is the number every other calculator reports.
- The expected value of one test is the same calculation with the lift replaced by win rate x average winning lift, because a lift you only get one time in five is worth one fifth as much.
- A programme adds tests per month x win rate x average winning lift to revenue per visitor each month. A winner landing in a given month is credited half that month and in full thereafter, so the lift in force by month m is that monthly amount multiplied by (m minus 0.5).
- The annual figure is the sum of those twelve monthly amounts, capped so the lift in force never exceeds 30%.
The programme maths adds lifts rather than compounding them, which is worth naming because it is the conservative choice. Multiplying (1 + lift) terms together gives a slightly larger answer than adding them. We add, so the projection understates rather than overstates, and so the whole thing stays reproducible in a spreadsheet with nothing more exotic than a MIN function.
What the model cannot know is whether your store is one where testing works. Traffic, order values and the quality of the hypotheses all move these numbers more than any parameter on this page. Treat the output as the shape of the return rather than a forecast, and if you want it read against your actual numbers rather than these, that is a conversation rather than a calculator.
Common questions
It depends entirely on your baseline. On a store doing 80,000 sessions a month at 2.4% conversion and a $78 average order value, monthly revenue is about $149,760, so a 5% lift on revenue per visitor is worth roughly $7,490 a month, or $89,900 a year if it holds. The important qualifier is "if it holds": that figure assumes the test wins, which most do not, so the expected value before you run it is considerably lower.
Because the per-test figure includes the probability that the test does not win. Other calculators multiply your hoped-for lift by your revenue and stop, which silently assumes every test succeeds. If roughly one test in five produces a shippable winner, the expected value of one test is about a fifth of the headline. The programme figure on the same page is larger than a naive annual number, because a year of testing is 24 tests rather than one.
Ours runs around one test in four or five, which is the 20% default here. Published figures across the industry range from about one in eight to one in three, and they are not comparable between teams: a programme that only tests bold, well-researched changes will show a higher rate on fewer tests than one churning through button colours. If you have your own historical rate, use it. It is the single input that moves this calculation most.
Each month a programme adds an expected lift equal to tests per month multiplied by win rate multiplied by average winning lift. A winner is credited half of the month it lands in and fully thereafter, so the lift in force by month m is that amount multiplied by m minus a half. The annual figure sums those twelve months. Lifts are added rather than compounded, which understates slightly, and the total is capped at 30% for the year.
Because compounding models break down past there and start producing numbers no store has ever delivered. A cap makes that explicit rather than letting the arithmetic run away, and it is a useful filter in the other direction too: if an agency shows you a projection above 30% annual uplift from testing alone, the model is doing the selling rather than the evidence.
It accounts for them by not counting them, which is the honest treatment of a losing test that never ships. The win rate input is what carries them: a 20% win rate means four in five tests contribute nothing to the revenue projection. It does not model a shipped change that quietly costs you money, because a competent programme catches those before they ship, and if yours does not, that is a QA problem rather than a forecasting one.
Put this calculator on your own site
Free to embed, no attribution deal and no tracking script. The honest version of this calculation is not available anywhere else, which is rather the point of linking to it.
<iframe src="https://optyv.com/tools/revenue-impact-calculator/embed" width="100%" height="900" style="border:0;max-width:760px" title="A/B test revenue impact calculator" loading="lazy"></iframe>
<p>Calculator by <a href="https://optyv.com/tools/revenue-impact-calculator">Optyv</a></p>The iframe is a fixed 900px tall in this version. On a narrow column the calculator scrolls inside the frame rather than clipping.
Quoting a figure?
If you are citing one of these numbers, this says which calculator produced it and on what assumptions.
<a href="https://optyv.com/tools/revenue-impact-calculator"><img src="https://optyv.com/badge/roi-calculated.svg" alt="Testing ROI calculated with Optyv" width="230" height="40"></a>Want to know which of these lifts your store can actually get?
The number on this page is the shape of the return, not a forecast. We will read it against your traffic, your templates and your order values, and tell you plainly if testing is the wrong tool at your scale.
Get a free audit ↗