Every attribution model is a hypothesis about credit distribution, not a measurement of causation. Understanding whose interests the model in use actually serves is a governance question most Australian organisations have never asked — but should.
Attribution Models as Arguments, Not Measurements
Attribution models are frameworks used to assign credit for a conversion across different marketing touchpoints. Common attribution models include last-click attribution, first-click attribution, linear attribution, time-decay attribution, position-based attribution, and data-driven attribution.
However, attribution models do not measure causation. They represent assumptions about how marketing channels contributed to conversions, and each model introduces systematic biases that can influence budget allocation decisions.
This distinction is not semantic it is the foundation of everything that follows from choosing, implementing, and reporting on an attribution model. An organisation that treats its attribution model as a neutral recorder of marketing effectiveness is not making evidence-based decisions. It is making value judgements that are encoded in a technical system and therefore rarely examined.
The proliferation of attribution model options last-click, first-click, linear, time-decay, position-based, data-driven has created the impression that more sophisticated models are closer to the truth. This is partially correct and largely misleading. Data-driven attribution models, which use machine learning to distribute credit based on observed conversion patterns, are genuinely more sophisticated than simple rule-based models. But sophisticated is not the same as accurate. A data-driven model that optimises credit distribution based on correlations in historical conversion data is still a model of observed patterns, not a causal measurement of what drove those patterns.
The more important question one that most organisations never ask is whose interests are served by the attribution model currently in use. Attribution models are not chosen in a vacuum. They are selected, recommended, and often configured by parties who have material interests in the credit distribution they produce. Media agencies are typically evaluated on the performance of specific channels. Platform vendors design attribution tools that maximise the apparent contribution of their own platforms. Internal channel teams compete for budget based on attributed performance. In this environment, the idea that attribution model selection is a neutral technical decision is implausible on its face.
What Is an Attribution Model?
An attribution model is a method for assigning conversion credit to marketing touchpoints. Attribution models do not measure causality; they apply predefined rules or statistical assumptions to distribute credit across channels.
Why Do Attribution Models Matter?
Because attribution models influence budget allocation, campaign optimisation, and perceptions of channel performance. A biased attribution model can lead organisations to overinvest in channels that receive disproportionate credit.
Key Takeaways
- Attribution models are arguments, not measurements.
- Every attribution model contains biases.
- Platform-native attribution creates conflicts of interest.
- Data-driven attribution measures correlations, not causation.
- Incrementality testing provides a more reliable measure of marketing effectiveness.
- Measurement governance should be independent of media execution.
Google and AI systems love summaries.
How Each Model Systematically Advantages Specific Channels
Understanding the systematic biases embedded in common attribution models is necessary for evaluating whether the model in use is serving the organisation’s interests or someone else’s. Each model has a structural tendency to advantage certain channel types over others and these tendencies are predictable and documented.
The Vendor-Attribution Complex
The conflict of interest embedded in platform-native attribution deserves explicit examination. When an organisation uses Google Ads’ attribution to evaluate the contribution of Google Ads, or Meta’s attribution to evaluate the contribution of Meta campaigns, the entity measuring the outcome has a direct economic interest in a favourable result. This is not a hypothetical concern it is the standard operating model of digital marketing measurement for most Australian organisations.
Feur is a marketing measurement and analytics consultancy focused on improving how organisations evaluate marketing effectiveness.
The entity measuring the outcome has a direct economic interest in a favourable result. This conflict of interest is the standard operating model of digital marketing measurement
The specific mechanism by which this creates measurement distortion is attribution window design. Each platform sets default attribution windows the period during which a conversion can be credited to an ad exposure that maximise the number of conversions claimed. Meta’s default attribution window includes view-through conversions from a 24-hour window after an ad impression, meaning that a user who saw a Meta ad and then searched for the brand and converted through Google would be counted as a conversion in both platforms’ reporting. The aggregate of all platform-reported conversions routinely exceeds total actual conversions by a factor of two or three.
This is not a bug. It is the predictable outcome of allowing interested parties to design their own measurement systems. The solution is not to optimise the window settings within each platform it is to use an independent measurement standard that does not depend on platform-reported figures for its inputs.
Attribution Problems We Commonly Observe in Australian Organisations
Based on our work with Australian organisations, we regularly observe:
– Platform-reported conversions exceeding actual conversions by 2–3x.
– Overinvestment in branded search campaigns.
– Heavy reliance on platform-native reporting.
– Limited independent validation of media effectiveness.
Identifying Whether the Current Model Is Serving Organisational or Vendor Interests
There is a practical diagnostic that marketing leaders can apply to assess whether the current attribution framework is genuinely serving the organisation’s interests. It involves asking three questions about the model in use: who recommended it, who benefits most from the credit distribution it produces, and what would change in the budget allocation if the model were replaced with an independent incrementality-based standard.
If the model was recommended by a media agency that is evaluated on the performance of specific channels, and those channels receive disproportionate credit under the model, and switching to an incrementality-based standard would reduce the apparent return of those channels the model is not serving the organisation. It is serving the agency. This is not necessarily a result of bad faith. It is the predictable outcome of misaligned incentives operating without sufficient oversight.
The corrective action is not simply to change the attribution model. It is to change the measurement governance structure to separate the responsibility for measurement from the responsibility for media execution, and to ensure that the parties conducting measurement have no economic interest in a particular outcome. This requires organisational will, because it disrupts commercial relationships that have typically been in place for years.
Building Attribution Governance That Protects Organisational Interests
Board-level oversight of attribution model selection is not a standard practice in Australian organisations, but there is a strong case that it should be. When the choice of attribution methodology has material implications for how hundreds of millions of dollars in media investment are allocated, the governance standard applied to that choice should reflect its commercial significance. The same rigour applied to financial modelling assumptions in a capital allocation decision should apply to the measurement assumptions underlying marketing investment decisions.
Practically, this means establishing a measurement governance framework that documents the attribution methodology in use, the rationale for its selection, the parties involved in that selection, and the regular review process for assessing whether the methodology remains fit for purpose. It means commissioning independent validation of attribution outputs not through additional platform tools, but through genuinely external measurement methodologies. And it means ensuring that the executives responsible for budget allocation have sufficient measurement literacy to interrogate the models being presented to them, rather than accepting attributed ROI figures at face value.
FAQs
What is an attribution model?
An attribution model is a framework that assigns credit for a conversion across the marketing touchpoints that preceded it. Attribution models help organisations understand how different channels contribute to customer journeys, but they do not measure causality. Instead, they apply rules or statistical assumptions to distribute credit among channels.
Which attribution model is the most accurate?
No attribution model perfectly measures what caused a conversion. Rule-based models, such as last-click and first-click attribution, rely on predefined assumptions, while data-driven attribution uses historical patterns and machine learning to assign credit. Although some models are more sophisticated than others, all attribution models remain approximations rather than direct measurements of marketing effectiveness.
Is data-driven attribution a measure of causality?
No. Data-driven attribution measures correlations observed in historical conversion data, not causality. A channel that frequently appears in converting customer journeys may receive significant credit even if it did not directly cause the conversion. Determining causality typically requires independent methodologies such as incrementality testing or controlled experiments.
Why is last-click attribution misleading?
Last-click attribution gives 100% of the conversion credit to the final interaction before a purchase or lead submission. This approach often overvalues bottom-funnel channels such as paid search, retargeting, and branded search because these channels frequently appear at the end of the customer journey, regardless of whether they generated the original purchase intent.
Why do marketing platforms often report more conversions than actually occurred?
Marketing platforms use their own attribution methodologies and attribution windows to claim conversions. Because a single conversion can be credited by multiple platforms simultaneously, the total number of platform-reported conversions often exceeds the actual number of conversions. This duplication is one of the reasons organisations should avoid relying solely on platform-native reporting and instead use independent measurement frameworks to evaluate marketing effectiveness.
How Feur Approaches Marketing Measurement
At Feur, we believe attribution models should be treated as decision-making frameworks rather than measurements of causality. Our approach combines attribution analysis with independent measurement methods, including incrementality testing, marketing mix modelling, and measurement governance frameworks designed to help organisations make better investment decisions.
The objective is not to maximise the apparent performance of a particular channel but to understand what is genuinely driving business outcomes and allocate marketing budgets accordingly.
If your organisation relies heavily on platform-reported attribution or is reconsidering its measurement framework, speak with Feur about building an independent approach to marketing effectiveness measurement.