Changing an attribution model changes how credit is assigned to the same observed journey. It does not change what the customer did. Use the model to answer a specific reporting question, then check how sensitive the answer is to its assumptions.
What a model actually decides
A buyer may encounter an ad, a search result, an email, and a return visit before converting. Attribution decides how to assign credit across those recorded touches.
A first-touch rule assigns credit to discovery; a last-touch rule assigns it to the final interaction. The underlying journey stays the same while the reporting view changes.
Common attribution approaches, at a glance
| Model | Who gets credit | Best for | Blind spot |
|---|---|---|---|
| First-touch | The first interaction | Demand gen, new markets | Ignores everything that closed the deal |
| Last-touch | The final interaction | Short, simple sales cycles | Over-credits brand and bottom-funnel |
| Linear | Every touch equally | Long, multi-touch journeys | Treats a glance like a sales demo |
| Time-decay | Recent touches more | Considered B2B purchases | Still under-credits early discovery |
| Data-driven | Model-based credit | When the platform’s documented model fits the question | Method and inputs differ by platform |
The asterisk nobody mentions
Some platforms supplement observed events with modeled key events when they cannot be observed directly. Google documents that modeled key events may appear in eligible GA4 reports.
That does not mean every reported touchpoint is modeled, or that attribution output proves a channel caused an outcome.
Stop hunting for the true model. Pick the lens that fits the decision, and be honest about what it can’t see.
How to actually choose
A framework that ends the religious war:
- Growing awareness or entering a new market? Lean first-touch.
- Short cycle, one or two touches? Last-touch is fine, and simpler.
- Long B2B journey you’re trying to understand? Linear or time-decay.
- Does a platform’s documented data-driven method fit the question? Compare it with a simpler rule and note the assumptions.
- Reporting to a CFO? Anchor everything to blended results and revenue.
The move that beats picking a model
A team can bring available touch data into one dataset and apply a documented set of rules. The resulting view still depends on source coverage and identity links; compare model outputs for diagnosis and label the selected reporting rule.
The guide to why platform numbers never match covers the wider issue. For help reviewing your setup, book a tracking audit. For an example of how that work is scoped, see the attribution audit scope of work example.