What is attribution?
Attribution assigns credit for a conversion to recorded marketing touchpoints. A person might click a search ad, later read an email, and then buy after a direct visit. An attribution model decides which of those interactions receives credit in a report.
The model is a reporting rule. It cannot observe every influence on a decision or prove that an ad caused the sale.
Common attribution models
| Model | How it assigns credit | What it helps you see |
|---|---|---|
| First touch | All credit goes to the first eligible recorded touchpoint. | Which channels start measured journeys. |
| Last touch | All credit goes to the last eligible recorded touchpoint. | Which channels appear closest to conversion. |
| Last non-direct | Credit goes to the latest eligible touchpoint other than a direct visit. | The preceding channel when someone returns directly. |
| Linear | Credit is split evenly across recorded touchpoints. | The channels present throughout a journey. |
| Position based | A chosen share goes to the first, last, and middle touchpoints. | How results change under stated weights. |
| Data driven | A model estimates credit from observed paths. | Patterns that a fixed rule may miss. |
Take a journey with a search ad, an email click, and a direct visit. First touch credits search. Last touch credits direct. Last non-direct credits email. None of these reports alone tells you what would have happened without the ad or email.
What changes the result?
An attribution window limits how far back a touchpoint can receive credit. A seven-day window may exclude a search click that a 30-day window includes. Platforms also differ in whether they count ad views, direct visits, and cross-device interactions.
Data coverage matters as much as the model. A blocked request, missing campaign tag, absent consent, or unlinked device can remove a touchpoint from the recorded path. Offline activity may never appear in a digital attribution report. Treat the output as credit across observed interactions.
How should teams use it?
Choose one conversion definition and document the model and window beside each report. Compare models to see how sensitive channel rankings are to the rule. Reconcile conversion totals with first-party orders or leads, then use holdout tests or other experiments when the decision depends on causal lift.
For a broader operating guide, read marketing attribution. For a causal test, read incrementality testing.
How Hardal handles it
Hardal's measurement product collects first-party events across web, mobile and server sources and sends the same records to each ad platform, so attribution models start from consistent inputs. Hardal also lets you define your own custom rules and custom weights.
Related terms: Data discrepancy, Custom rule, Custom weight, Web-to-app