Key takeaways
- Attribution allocates observed credit; it does not by itself estimate the causal effect of media.
- Document eligible channels, identity scope, lookback windows, event time, and model changes.
- Reconcile attributed conversions with raw events and CRM outcomes before discussing channel quality.
- Use experiments, MMM, qualitative evidence, and operational constraints alongside attribution.
01
Name the question the report actually answers
Last click, data-driven attribution, and an external model can assign different credit to the same observed conversion. Google Analytics documents that its reporting attribution setting affects conversion and revenue reports while leaving user and session data unchanged. The model is a reporting rule applied to paths, not a rewrite of what the person did.
Start every decision memo with a plain statement: “This report allocates observed conversion credit among eligible tracked interactions under this lookback window.” That sentence prevents attributed revenue from quietly becoming incremental revenue.
Swipe to compare every column
| Layer | What it represents | Do not call it |
|---|---|---|
| Raw event | A recorded action at a stated time | Causal impact |
| Attributed event | Credit assigned under a model | Ground truth channel value |
| CRM outcome | A lifecycle or commercial state | Automatically matched media effect |
| Experiment or causal model | Estimated difference from a counterfactual | A complete journey narrative |
02
Create an attribution contract
Record the property, eligible channel scope, conversion actions, identity stitching, consent behavior, lookback windows, timezone, interaction or conversion date, direct-traffic treatment, and attribution model. Keep the effective date of every change.
This contract matters because a configuration change can alter historical and future reporting. Without a version log, a trend line may compare two accounting rules while looking like one continuous measure of performance.
03
Reconcile before interpreting
Compare raw conversion events, attributed conversions, imported offline outcomes, and CRM records by stable event or cohort identifiers. Explain timing differences and excluded records. A clean bridge table is more useful than forcing every system to produce the same daily total.
Then inspect path coverage: missing UTMs, self-referrals, cross-domain breaks, unconsented sessions, offline interactions, deleted cookies, and marketplace purchases. Attribution cannot credit an interaction the measurement system never observed.
04
Use a portfolio of evidence for budget decisions
Attribution is valuable for consistent reporting, path diagnostics, and operational feedback. Randomized experiments can estimate lift for bounded treatments. MMM can examine aggregate channel effects and response curves. Sales notes and customer research can reveal influences that neither browser tracking nor media logs capture.
Let disagreements become questions. If attributed performance rises while qualified pipeline falls, inspect event definitions, channel mix, and sales follow-up. If an experiment contradicts platform ROAS, calibrate the planning model instead of selecting the friendliest number.
Primary sources and further reading
Use the source material to validate details against your own context and current platform configuration.
- Google Analytics Admin API: Attribution settings
- Google Analytics Data API: Conversion reporting basics
- Google Research: Near Impressions for Observational Causal Ad Impact
This field note follows the XenGrowth editorial policy: primary sources where available, visible limitations, material review dates, and no invented first-hand experience.
Stay with the problem
