What question should attribution answer?
Begin with a decision, not a model. Budget allocation, campaign diagnosis, creative learning, and customer-journey analysis require different grains and evidence. A single dashboard cannot make one attribution number equally suitable for all four.
State the outcome, population, reporting window, and decision owner. Then identify which interactions are observable and which are missing because of consent, device changes, offline activity, platform boundaries, or incomplete identifiers.
- Outcome: key event, qualified lead, customer, revenue, or another agreed business state.
- Scope: channel, campaign, product, market, device, or customer segment.
- Window: when interactions and outcomes are eligible for credit.
- Decision: the action that the analysis can approve, reject, or investigate.
Which attribution evidence should a team compare?
Platform, analytics, and CRM reports can all be valid within their own definitions while producing different totals. Keep those definitions visible. Do not merge platform-attributed conversions and call the sum unique customers.
| Evidence layer | What it describes | Useful for | Main limitation |
|---|---|---|---|
| Platform attribution | Outcomes credited by one advertising platform | Campaign delivery and in-platform optimization review | Overlapping credit across platforms |
| Analytics attribution | Measured site or app paths under an analytics model | Cross-channel journey and landing-page analysis | Consent, identity, and tagging coverage |
| CRM outcomes | Recorded lead, pipeline, and customer states | Quality and commercial outcome review | Source mapping and sales-process hygiene |
| Controlled experiment | Difference under an intentional comparison | Testing causal impact of a material change | Eligibility, duration, interference, and cost |
How should models and windows be documented?
Always show the attribution model, interaction scope, and lookback window beside the result. Google Ads and Google Analytics expose attribution settings and reports, but their outputs answer product-specific questions. Model changes can move reported credit without changing underlying customer behavior.
- 1
Record the source product and report name.
- 2
Record model, lookback window, conversion definition, and time zone.
- 3
Annotate tagging, consent, import, or identity changes.
- 4
Compare trends within a stable definition before comparing totals across systems.
- 5
Reconcile against business outcomes at an agreed cadence.
When should attribution influence budget?
Use attribution as one input alongside incrementality evidence, lead quality, margin, inventory, sales capacity, and strategic coverage. A channel that appears late in a measured path may be harvesting demand; an upper-funnel channel may be under-observed. The safest budget changes are bounded, reversible, and monitored against a pre-agreed outcome.
- Prefer a controlled test when the budget decision is material and testable.
- Use triangulation when a clean experiment is not feasible.
- Avoid reallocating on a short-term model movement alone.
- Keep brand, acquisition, retention, and market-entry objectives distinct.
What does a practical attribution review look like?
A monthly review can compare source-reported outcomes, analytics key events, CRM-qualified leads, and controlled-test evidence. The meeting should explain differences rather than forcing agreement. Decisions should cite the evidence layer they rely on.
Evidence
Official sources
These primary references support the platform-specific definitions used in this Guide. Growomo commentary and workflows remain distinct from provider documentation.