Free tool
Test Bandwidth Calculator
Work out how sensitive each of your test areas actually is, and how many tests a year your traffic will support. For every area the calculator finds the smallest uplift you could reliably detect in the time you have, which is usually the number that decides whether testing there is worth doing at all.
Chance of spotting a real effect.
Two-tailed unless a drop would genuinely not matter.
Applied to new rows. Change any row on its own.
Drop it to allow for freezes and peak periods.
Your test areas
Swipe the table sideways to reach the detectable uplift.
| Test area | Weekly visitors |
Weekly conversions |
Weeks | Traffic % | Variations | Confidence | Smallest uplift you could detect |
Tests / year |
|---|
Reading the verdict.
The uplift is always a positive number, so the colour sits on the word beside it rather than
on the figure. Sensitive means 5% or less, small enough that a realistic change could
move it. Workable is 5% to 10%, detectable but only for a change with real force behind
it. Too blunt is above 10%: very few genuine changes shift a metric that far, so a flat
result there would tell you almost nothing.
How the detectable uplift is worked out. Your available sample is weekly visitors times weeks times the share of traffic in the test. That is split between the variations, and the calculator searches for the smallest relative uplift a two-proportion test could detect on that sample at your confidence and power. Tests per year is the testing weeks divided by the weeks per test, divided again by the traffic share, since a test on half the traffic leaves room for another one beside it. That assumes you can run them side by side without them interfering, which is worth checking before you promise the number to anyone.
How the detectable uplift is worked out. Your available sample is weekly visitors times weeks times the share of traffic in the test. That is split between the variations, and the calculator searches for the smallest relative uplift a two-proportion test could detect on that sample at your confidence and power. Tests per year is the testing weeks divided by the weeks per test, divided again by the traffic share, since a test on half the traffic leaves room for another one beside it. That assumes you can run them side by side without them interfering, which is worth checking before you promise the number to anyone.
Share and embed
Use this tool on your own site
Free to embed, no attribution required, though a credit link is appreciated and included by default. The tool runs entirely in the visitor’s browser, so nothing you enter is sent anywhere.
px
