Same conversion. Six different "winning" channels.
Enter one real conversion path and watch a channel go from getting nearly all the credit to nearly none of it, depending only on which model produced the report.
Your numbers
How fast credit decays going back in time. 7 days is a common default.
Touchpoints, in order
| Channel | Linear | First | Last | U-Shape | W-Shape | Decay |
|---|---|---|---|---|---|---|
| Paid Social | 25% | 100% | 0% | 40% | 30% | 6% |
| Organic Search | 25% | 0% | 0% | 10% | 30% | 15% |
| 25% | 0% | 0% | 10% | 10% | 30% | |
| Direct | 25% | 0% | 100% | 40% | 30% | 49% |
"Paid Social" gets anywhere from the smallest to the largest share of credit depending only on which model you pick. That's a swing of 100 percentage points on the exact same conversion path. That's the actual argument for picking one model and reporting it consistently, not switching models until the number matches what a channel owner wants to hear.
No model here is "correct". Each one answers a different implicit question about which touch mattered. Time-decay and U-shaped are the most commonly defaulted-to in practice because they avoid first-touch and last-touch's all-or-nothing extremes.
Calculated in your browser. Nothing is sent anywhere, and nothing is stored.
The maths, so it isn't a black box.
It runs one path through six real models
Linear, first-touch, last-touch, U-shaped (position-based), W-shaped, and time-decay: the standard credit-splitting models most attribution platforms offer, applied to the exact same ordered touchpoints.
Time-decay uses a real half-life formula
Weight = 2^(−days-before-conversion ÷ half-life), normalized so credit sums to 100%: the same exponential-decay math attribution platforms use, with an adjustable half-life instead of a hidden default.
And flags the channel with the biggest model-dependent swing
The channel whose credit share moves most across the six models is usually the one an internal budget argument is actually about. Worth knowing before the debate starts, not after.
Questions about this calculation
There's no universally correct answer. Each model encodes a different assumption about which touch mattered, and none of them are causal. The practical discipline is picking one, understanding what it assumes, and reporting it consistently instead of switching models until a number looks better.
Terms used here
Other calculators
Where we do this work
We'll build this model on your real data.
A free 30-minute teardown where we build your contribution margin model from actual figures and show you where the funnel leaks. Yours to keep either way.