UNDERSTAND PATTERNS
Compare evidence before and after a change.
Build a before-and-after review sheet with explicit windows, coverage, counts and uncertainty.
In this guide 2 sections
Use the evidence carefully
#From a saved investigation choose Compare this page, then select separate before and after UTC ranges and the relevant layout groups. Read the sample counts and missing coverage alongside each percentage. A layout or traffic change can alter the population.
Comparisons are descriptive. They do not establish causation, recovered revenue, conversion lift or that a fix worked for every visitor. Collect suitable new eligible visits under your chosen recording mode if a period has insufficient observations; do not treat missing data as zero friction.
Keep a before-and-after review sheet
#| Record on the sheet | Why it matters |
|---|---|
| Question and unchanged destination | Keeps the comparison focused on a specific observation. |
| Edit time and separate UTC windows | Keeps deployment activity out of the intended periods. |
| Capture settings, filters, layout and counts | Exposes changes in coverage or population. |
| Replay references and unclassifiable cases | Lets another authorised reviewer check the interpretation. |
| Finding, uncertainty and next check | Allows an inconclusive result instead of an unsupported success claim. |
Use the comparison reference for operating instructions and the menu-rename example for a worked sheet. This worksheet is not an A/B testing tool.