Campaign Attribution Model
Definition
Campaign Attribution Model
A campaign attribution model is the rule that decides which of the touchpoints on a buyer’s path receive credit for a conversion, and in what proportion. Every model is a convention, not a measurement — none of them observes causation directly.
This glossary also carries a broader entry on attribution modeling as a discipline. This page covers the same mechanics applied to a specific campaign, and the two should be read together rather than as rivals.
The choice of rule changes the answer, sometimes dramatically. A last-click rule flatters search and retargeting; a first-touch rule flatters awareness work; neither is describing what actually persuaded anybody.
The honest posture is to fix one model, apply it consistently, and treat the output as a comparison device between campaigns rather than as a statement of truth about any one of them.
Agencies should not own the model. Asking a supplier to choose the rule that grades its own work is a conflict nobody needs to accept, however good the relationship is.
Key takeaways
- The model is a credit-assignment convention chosen in advance, not a discovered fact.
- Different rules produce different winners from identical underlying data.
- Consistency over time matters far more than picking the theoretically best rule.
- Offline, dark-social and word-of-mouth touches stay invisible to every model.
How it works
Each conversion carries a path of recorded touches. The model distributes a fixed amount of credit across that path, and the distribution rule is the only thing that varies between models.
Data collection limits everything downstream. Touches that were never recorded cannot receive credit, so a model’s output always overstates the channels that happen to be measurable.
A lookback window has to be set, and it is always arbitrary. Thirty days suits a short consumer purchase and hides most of what happened in a nine-month business sale.
The window should be derived from observed path lengths rather than from a platform default. Cross-device identity is the other silent limit — a buyer who researches on a phone and converts on a laptop looks like two people.
Measurement discipline is treated as a standing obligation in public digital work. The UK government’s measuring success guidance covers “measuring, reporting, analytics tools and techniques” as part of running a service.
| Model | Credit rule | Bias it introduces |
|---|---|---|
| Last click | All credit to the final touch | Favours search and retargeting |
| First touch | All credit to the first touch | Favours awareness and content |
| Linear | Equal credit across every touch | Favours high-volume channels |
| Time decay | More credit nearer the conversion | Favours late-stage activity |
Claims made in the campaigns being measured carry their own obligations. The Federal Trade Commission’s advertising guidance tells advertisers to “disclose the details of the deal up front”, whatever the medium.
Examples
Attribution arguments are usually budget arguments in disguise, and the model chosen quietly decides who wins them. The three cases below show how that plays out in practice.
A provider switches from last click to time decay. Its revenue per lead figures for content fall, and the budget conversation changes overnight.
A software firm runs two models side by side deliberately. The marketing operations manager reports both, which removes the temptation to pick the flattering one after the fact.
A retailer finds a channel invisible to every model. Only a holdout test, where the channel is switched off in one region, establishes what it was contributing.
Related terms
Attribution feeds budgeting, forecasting and agency evaluation, so the surrounding entries are mostly measures. Each one below is affected by the model choice rather than defining it.
- Digital marketing outsourcing: the arrangement whose performance the model judges.
- Pipeline velocity: speed through the funnel, not credit for entering it.
- Campaign manager: the role whose results the model reallocates.
- Advertising cost of sales: the ratio that moves when credit moves.
FAQ
Which model is most accurate?
None of them is accurate in the causal sense. Time decay and position-based rules tend to be least misleading for longer purchases, but all of them are conventions.
How often should the model change?
Rarely. Changing it breaks comparison with every prior period, so a change should be planned, announced — and run in parallel with the old model for a full quarter.
Can attribution prove what caused a sale?
No. Only an experiment, such as switching a channel off in one region, produces causal evidence. Attribution allocates credit within recorded data.
What about offline touchpoints?
They are usually missing entirely. Where they matter, a media mix model or a holdout test is a better tool than any touch-level attribution rule.
Does more data fix the problem?
It narrows the gap without closing it. The unrecorded conversation that actually decided the purchase remains unrecorded however much else is captured.
Who should choose the model?
Finance and marketing together. A model chosen by whoever it flatters will be argued about at every budget round thereafter.
Explore more marketing measurement guidance at Outsource Accelerator.
Independent
