An attribution model is the rule that splits credit for a sale between the touchpoints that led to it. The three most common models are simple, transparent and easy to compare.
A worked example
A customer's path before a $120 order:
- Hears a podcast ad (day 1)
- Clicks an Instagram ad (day 4)
- Searches the brand on Google and buys (day 6)
Here is how each model credits that $120:
- First touch: podcast $120, Instagram $0, Google $0
- Last touch: podcast $0, Instagram $0, Google $120
- Linear: $40 each
Same sale, three very different stories.
Last-touch attribution
Last touch gives all the credit to the final touchpoint before the sale. It answers: what finally got people to buy?
- Good for: retargeting, branded search and promotions that convert existing demand.
- Weakness: it hides the channels that introduced the customer. Upper-funnel channels like podcasts, video and TV look worthless under last touch, even when they started most journeys.
First-touch attribution
First touch gives all the credit to the first touchpoint. It answers: where do new customers come from?
- Good for: judging awareness channels and new-audience prospecting.
- Weakness: it ignores everything that happened after the introduction, including the channels that did the hard work of converting.
Linear attribution
Linear splits credit evenly across every touchpoint in the path. It answers: which channels show up in buying journeys at all?
- Good for: a balanced view when several channels work together.
- Weakness: it treats a quick glance and a decisive click as equal.
How to choose
Do not pick one model and stop looking. Switch between them on the same data:
- Start with linear as your default view.
- Check first touch before cutting an awareness channel, since it may be starting journeys that other channels finish.
- Check last touch before scaling retargeting, since it may be taking credit for sales that would have happened anyway.
In Trakinit you can switch models on every report, including two more: U-shaped (40% to the first and last touch, 20% across the middle) and time decay (a touch a week earlier counts half as much). Starter includes last touch; Growth and Enterprise include all five. See pricing.
Models need good matches first
A model can only split credit between touchpoints it has matched to the sale. The quality of those matches, through click IDs, hashed emails and networks, matters more than which model you choose. And the lookback window decides how far back a touchpoint can count.
Frequently asked questions
Is data-driven attribution better than rule-based models?
Data-driven models can be powerful but are hard to inspect. Rule-based models like first touch, last touch and linear are transparent, so you always know why a channel got credit. Many teams use rule-based models side by side.
Which model does Google Analytics use?
Analytics tools change their defaults over time and only see your website. An independent attribution tool applies one model consistently across every channel and store, including sales off your site.
Can I change models without re-collecting data?
Yes. Models are applied to the same matched touchpoints, so switching is instant.