Testing measurement before claims
Keep message transport, recipient response, and downstream action separate. Audience assignment, timing, creative, and missing data can confound a comparison.
Putting the guide into practice
Define a comparable audience and observation window, record costs and outcome sources, and preserve the analysis method before examining results.
- Question
- Assignment
- Outcome
- Analysis plan
- Limitations
Limits and review decisions
Do not publish performance lifts from an uncontrolled demonstration or a selectively chosen send. Report uncertainty and the conditions that could change the result.