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

From observation to action
StepQuestion
OBSERVEWhat changed, compared with what?
DIAGNOSEWhere and for whom did it change?
INTERPRETWhich explanations fit, and what evidence is missing?
DECIDEWhich action has the best expected value under the constraints?
ACTWho will do what, by when?
MEASURE AGAINWhat result would continue, revise or reverse the action?

Turn metric movements into questions

Five observations and the next investigation
ObservationDo not assumeInvestigate
Traffic fellDemand fellSource, visibility, tagging, season and landing availability
CAC increasedMedia became worseCost, conversion, qualification, mix and maturity
ROAS improvedProfit improvedSpend level, margin, incrementality and customer mix
Conversion fellThe site brokeTraffic quality, device, offer, stock and journey errors
Retention weakenedCRM failedCohort mix, product cycle, experience and measurement window

Ask six diagnostic questions

  1. What happened?
  2. Where did it happen?
  3. Why might it have happened?
  4. What evidence supports that explanation?
  5. What should we do next?
  6. 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

  1. 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.

  2. 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.

  3. 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 ↗