---
name: "creative-test-readout"
description: "Read an ad test by sample size and spread, and say whether there is a winner yet."
triggers: ["ad test results", "A/B test ads", "creative test", "which ad won", "test readout"]
version: "1"
---
# Creative test readout
Use this for one test between two or more ad variants.
## Check the setup
State what was tested, the one idea that differs, the dates, the audience,
and the measure agreed before the test began. Check that variants got a fair
share of spend, ran over the same days, and pointed to the same page. If the
setup was uneven, say so first. For more than two variants, name the one
comparison set before the test, or choose a method that accounts for all
comparisons.
## Read the result
For each variant show impressions, clicks, conversions, spend, and the rate
being judged. Give the gap between variants and whether it is larger than
the noise at this sample size. Use a stated method and show it. Before calling
a winner, state the baseline rate, minimum effect worth detecting, power, and
confidence level used to estimate the required sample.
For more than two variants, call a winner only for the comparison set before
the test, or after using and showing a test that accounts for all comparisons.
Say one of:
- **clear winner**, with the margin and sample;
- **leaning**, not yet safe to act on;
- **not enough data**, with the sample still needed and how many days that
takes at current spend.
## Decision and next test
Recommend keep running, stop, or roll out, and the next idea worth testing.
Do not pause, scale, or edit ads yourself; recommend for approval. Never call
a winner early or invent figures to fill a thin test.