THE SHORT ANSWER
Turn marketing data into a decision through OBSERVE → DIAGNOSE → INTERPRET → DECIDE → ACT → MEASURE AGAIN. Separate what happened from why it may have happened, test competing explanations, include business constraints and record how the action will be evaluated.
Use the decision loop
| Step | Question |
|---|---|
| OBSERVE | What changed, compared with what? |
| DIAGNOSE | Where and for whom did it change? |
| INTERPRET | Which explanations fit, and what evidence is missing? |
| DECIDE | Which action has the best expected value under the constraints? |
| ACT | Who will do what, by when? |
| MEASURE AGAIN | What result would continue, revise or reverse the action? |
Turn metric movements into questions
| Observation | Do not assume | Investigate |
|---|---|---|
| Traffic fell | Demand fell | Source, visibility, tagging, season and landing availability |
| CAC increased | Media became worse | Cost, conversion, qualification, mix and maturity |
| ROAS improved | Profit improved | Spend level, margin, incrementality and customer mix |
| Conversion fell | The site broke | Traffic quality, device, offer, stock and journey errors |
| Retention weakened | CRM failed | Cohort mix, product cycle, experience and measurement window |
Ask six diagnostic questions
- What happened?
- Where did it happen?
- Why might it have happened?
- What evidence supports that explanation?
- What should we do next?
- How will we know if it worked?
Keep a competing explanation alive until the evidence distinguishes it. A tracking change and a customer change can produce the same chart shape and require entirely different actions.
Evidence & context: Google Analytics Help
Write a decision record
Record the observation, scope, chosen explanation, alternatives, action, owner, expected effect, guardrails and review date. This turns analysis into an organizational memory and makes later learning possible.
For campaign-specific diagnosis, use How to Diagnose an Underperforming Campaign.
Evidence & context: Google Research
Sources & further reading
- Data differences between reports and explorations
Google Analytics Help. Official explanation of differences caused by supported fields, filtering, retention, thresholds, modeling and processing. It covers GA4 surfaces, not every cross-platform discrepancy.
- API dimensions and metrics
Google Analytics. Official GA4 reporting definitions checked 13 September 2026, including session source, medium, referral and landing-page dimensions. Attribution remains limited to observable interactions.
- Methods for Measuring Brand Lift of Online Ads
Google Research. Original research using randomised experiments to estimate advertising effects; no universal lift or ROI benchmark is inferred.
Examples and exercises are illustrative unless attributed to a source. No independent expert review is claimed.
A correction, a counterexample or an experience worth sharing?
Join the conversation ↗