Experimentation
Run length plan
Know before launch whether a test can realistically answer its question, and roughly how long it will need to.
Most disappointing tests were never going to produce an answer. Either they were stopped the moment the numbers looked interesting, or they ran for months on too little traffic to detect the change they were testing. The Run length plan is there to catch that before launch rather than after it.
A short wizard asks three things: how much traffic the test will get, the original's conversion rate, and the smallest lift worth acting on. It answers with an estimated run length, such as “est. 4 weeks”. The label is deliberate: it is an estimate, not a promise, and the product says so.
The starting figures come from your own data: the traffic your past experiments received, your website's traffic as Pertento measures it, and your past conversion rates. Where you know better, type your own figures in instead.
Every option shows the trade-off, because a smaller lift needs a longer test. Seeing the two side by side is usually what decides whether an idea is worth running as it stands. The plan is saved with the experiment and shown in its schedule and on the Library timeline.