How the five models split the credit
- First touch gives the whole order to the first touchpoint of the journey.
- Last click gives it to the last touchpoint before the order.
- Linear gives every touchpoint the same share.
- Time decay weighs each touchpoint by 0.5 to the power of (days before the order ÷ half-life), then scales the weights to 100%. With a half-life of 7 days, as in Google's model, a touchpoint 7 days before the order counts half as much as one on the day of the order.
- Position-based gives 40% to the first touchpoint, 40% to the last, and splits 20% among those in between. With two touchpoints, the simulator splits 50/50.
Example. A €120.00 order after four touchpoints, the journey the simulator starts with:
| Touchpoint | First touch | Last click | Linear | Time decay | Position-based |
|---|---|---|---|---|---|
| Meta Ads, 18 days before the order | €120.00 (100%) | €0.00 (0%) | €30.00 (25%) | €8.42 (7%) | €48.00 (40%) |
| Google Ads (non-brand), 9 days before the order | €0.00 (0%) | €0.00 (0%) | €30.00 (25%) | €20.52 (17%) | €12.00 (10%) |
| Email, 2 days before the order | €0.00 (0%) | €0.00 (0%) | €30.00 (25%) | €41.04 (34%) | €12.00 (10%) |
| Google Ads (brand search), on the day of the order | €0.00 (0%) | €120.00 (100%) | €30.00 (25%) | €50.02 (42%) | €48.00 (40%) |
The same order gives Meta Ads anything from €0.00 to €120.00, and brand search anything from €0.00 to €120.00. Nothing about the channels changed, only the rule.
How to read the simulator's result
A channel whose credit stays about the same from model to model holds the same place in the journey whichever rule you apply. One whose credit swings, like Meta in the example, depends on whether it opens the journey, closes it or sits in between. Rebuild the journeys your customers actually take, with their real gaps in days, and you see which of your channels the choice of model moves most.
Which model to judge each channel on is the attribution model entry's view on which model fits which decision. How far first and last touch move the median ROAS of the D2C brands in our panel, the ROAS calculator shows next to your own ROAS.
From the simulator to adtribute's models
The simulator applies each rule to one journey you type in. In adtribute, a model applies it to every order, from the touchpoints your tracking recorded (docs: Attribution models in adtribute).
- First touch, last click, linear and position-based are standard models in adtribute: First Touch, Last Touch, Linear and Position Based (40/20/40).
- Time decay is not a standard model, but a model of your own can distribute by time decay.
Brands with many channels can go further and build models of their own. SNOCKS has eleven attribution models, five of them custom models built with adtribute, and put the reason this way: “Each channel wants to look best with its attribution model... With an attribution model, you assign a single sale to different marketing touchpoints. This helps you understand which marketing channel really had an impact.” (SNOCKS case study)
The attribution model entry has the six standard models, the building blocks every model combines, from the window to the fallback model, and where you choose between conversion date and click date.
Traps when you compare attribution models
- Direct visits. Counted as a touchpoint, a direct visit right before the order takes all of a last-click model's credit. Toggle the option in the simulator to see the difference.
- The window. A 7-day window drops every touchpoint before it: a journey of three weeks loses the channel that started it.
Related glossary entries, market data and docs
- Attribution model, attribution window and touchpoint
- First-touch attribution and last-click attribution: the two single-touch models in detail
- Good ROAS for Meta (Facebook) Ads: Last Touch vs First Touch
- Good ROAS for Google Ads (Non-Brand): Last Touch vs First Touch
- Attribution models in adtribute (docs)
- ROAS calculator and UTM builder
Attribution model simulator: frequently asked questions
Which half-life should I set for time decay?
Start with 7 days, the half-life of Google's time decay model as Search Ads 360 documents it: a touchpoint 7 days before the order then counts half as much as one on the day of the order. Shorter half-lives suit short, promotional journeys and move the split towards last click. Longer ones suit considered purchases and move it towards linear.
Can we give our own model a fixed 7-day window?
Yes: in adtribute, the window is one of every model's settings, so a model of your own can count only the touchpoints of the 7 days before each conversion (docs: Attribution models in adtribute, Build a model). The window field here does the same to one journey.
What happens when a channel appears twice in a journey?
Each touchpoint gets its own share, and the channel gets the sum of its touchpoints. Under linear, a channel with two of four touchpoints gets half the order; under time decay, its more recent touchpoint counts more. First touch and last click still credit one touchpoint each, so a repeat changes their result only when it is the first or the last touchpoint.
How does position-based split a journey with only two touchpoints?
Half each: with nothing in between, the first and the last touchpoint get 50% each. From three touchpoints on, position-based gives 40% each to the first and the last and splits the other 20% among those in between. With one touchpoint, every model gives it the whole order.