How the money is calculated
Monthly revenue is visitors × conversion rate × average order value. A lift is applied relatively, which is the distinction most people get wrong: a 10% lift on a 2% conversion rate is 2.2%, not 12%. Extra monthly revenue is simply the difference, and the annual figure is that twelve times over — it assumes the improvement holds and your traffic does not change, which is an assumption, not a promise.
How “to prove it” is calculated
The visitor count is a standard two-proportion sample size at 95% confidence and 80% power, computed from your own baseline rate and the size of the effect you are looking for. Days assume an even split between the two versions, using your traffic. Smaller effects need dramatically more traffic: on a 2% baseline, detecting +20% takes about a fifth of what +10% takes.
Past 90 days a row is marked out of reach rather than given a longer estimate. That is not pessimism — it is that a test running through a season change, a price change and three months of cookie churn is no longer comparing two versions of the same store.
These are the same functions that size a real experiment inside Looplift. The public page does not get friendlier statistics than the dashboard does.
What the numbers do not say
- They are not a forecast. The table says what a lift would be worth. It does not say you will get one — most individual experiments do not win, which is why the practice is a portfolio rather than a purchase.
- A win is not permanent. A lift measured over four weeks in one season is evidence about those four weeks. It is strong evidence — it just is not a contract with next year.
- Measurement loses some events. Consent tooling, ad blockers and cross-device journeys cost every browser-side measurement some data. The defensible claim is that the loss falls on both versions alike, so the comparison holds — not that the counts are absolutely accurate.
If a row you want says out of reach
You have three real options, and buying a testing tool is not one of them. Test bigger changes, since large effects need far less traffic than small ones. Test earlier in the funnel, where the rates are higher and the samples fill faster. Or accept that at your traffic the honest method is evidence rather than proof — apply what session recordings, reviews and support tickets already tell you, and come back to testing when the store can settle a question in weeks.
More on the practice in what is CRO, and on why watching a live scoreboard produces false winners in why most A/B tests lie to you.