AIWorkDex
Marketing Practice

Make a campaign decision from channel evidence

Produce a traceable stop, change, or continue decision from configured GA4 evidence, with definitions, data gaps, and attribution limits attached.

Preparation
One measurement audit and one focused review session.
Difficulty
Advanced
Task
Review channel evidence
Back to the Task: Review channel evidence

Before you start

Get what you need before you start

  • Write the exact campaign decision before opening any dashboard.
  • Identify the implemented event that represents useful progress and the owner who can verify it.
  • Gather campaign dates, source tags, change logs, consent settings, platform definitions, and known incidents.
  • Use GA4 only for its documented source and key-event evidence; do not ask AI to reconcile platforms or infer causality.

Steps

Work through the method

Use the listed inputs and tools. Check the evidence when you need to verify a step.

  1. Define the decision and important event

    State the stop, change, or continue decision, campaign scope, audience, dates, important event, expected data sources, comparison, exclusions, and decision owner.

    Why it matters: A decision-first brief prevents the review from drifting towards whichever dashboard movement looks most persuasive.

    Input
    The campaign plan, business context, event implementation, baseline, budget, and operating constraints.
    Output
    A signed review brief with one decision, one important event, scope, comparison, exclusions, and owner.
    Tools in this stepGoogle Sheets

    Evidence for this step

  2. Audit collection and definitions

    Verify the event fires as intended, key-event marking, source tags, campaign dimensions, date range, attribution setting, consent route, exclusions, time zone, and material configuration changes.

    Why it matters: A combined report is misleading when the event, tag, consent, or definition changed during the comparison.

    Input
    The implementation, test events, GA4 property, campaign URLs, consent design, change log, and review brief.
    Output
    A measurement audit with passed checks, gaps, changed definitions, incidents, and a proceed or repair decision.
    Tools in this stepGoogle AnalyticsGoogle Sheets

    Evidence for this step

  3. Extract the bounded evidence

    Review the selected source, medium, campaign, date, and key-event views. Export the relevant rows with definitions and compare them with spend, delivery, site changes, outages, and other known factors.

    Why it matters: Traceable rows and surrounding context expose where a neat channel story exceeds the available evidence.

    Input
    The passed measurement audit, GA4 acquisition report, key-event report, platform delivery data, spend, and change log.
    Output
    A dated evidence table with source exports, definitions, observations, gaps, and competing explanations.
    Tools in this stepGoogle AnalyticsGoogle Sheets

    Evidence for this step

  4. Record the decision and limits

    State what changed, what remained unknown, attribution limits, plausible non-marketing causes, and the stop, change, or continue action. Assign an owner and review date.

    Why it matters: A useful campaign decision can acknowledge uncertainty without turning attributed events into a causal explanation.

    Input
    The review brief, audit, evidence table, campaign cost, operational context, and decision criteria.
    Output
    A signed decision record with evidence links, limits, competing explanations, action, owner, and follow-up date.
    Tools in this stepGoogle Sheets

    Evidence for this step

Success checks

Check the result before you finish

  • The review starts from one written campaign decision and one verified important event.
  • Event collection, tags, consent, dates, definitions, attribution, and configuration changes are recorded.
  • Every observation links to a dated source export and keeps data gaps visible.
  • The decision avoids cross-platform reconciliation, unsupported AI explanation, and causal claims.

Failure modes

Watch for these problems

  • The key event was marked after the review period or changed definition. Rebuild a comparable window or record the gap.
  • Campaign tags are missing or inconsistent. Repair tagging before attributing the observation to a named campaign.
  • Separate platforms use different event definitions. Keep them separate and compare only documented meanings.
  • A conversion decline is assigned to marketing alone. Add product, pricing, stock, site, season, and operational explanations.

Tools

Choose the tools you need

Sources

Read the sources behind this practice

Check what each source supports and where the advice has limits.