Before and after windows
Before and after windows are the main result lens.
The before window shows the baseline. The after window shows what happened after the change had time to matter.
Before and after windows result lens
Use windows that match the test.
If prompts run daily, a few days may be enough for a first read, but not always enough for a final decision. If the change is a page launch, give the page time to be found.
Comparison window
Use a before period and an after period that match the question set and timing of the test.
Evidence comparison
Compare the same types of evidence in both windows.
| Evidence | What to compare |
|---|---|
| Mentions | Did your brand appear more often? |
| Answer text | Did the explanation improve? |
| Sources | Did useful pages get read more often? |
| Competitors | Did competitor framing change? |
Movement that matters
Not every movement matters.
A small change in one answer may be noise. Repeated movement across the experiment question set is more useful.
Decision
Use the window comparison to make a practical decision.
Keep the change if the signal is useful. Adjust it if the signal is partial. Stop it if the result is unclear or harmful.