Creative and testing
Sample size
How many people, clicks or conversions a test needs before its result can be trusted.
What it is
It depends on the baseline rate, the smallest difference worth detecting and how sure the team wants to be.
Why it matters
For lead generation and e-commerceSmall samples produce large, unreliable differences. Many ad tests are stopped long before they have enough data.
In practice
Calculate it before the test with a standard calculator, and do not read the result until it is reached.
A common mistake
Choosing a test length by the calendar (one week) instead of by the data needed.
Read the evidenceThe winner is smaller than it looks All terms