For more than two decades, attribution has been the foundation of digital marketing measurement. It has shaped how organizations evaluate performance, influenced media investment and helped explain how customers move from their first interaction with a brand to conversion.
Entire reporting infrastructures have been built around attribution models, and in many organizations attribution data continues to influence significant marketing investment around the world.
Yet despite its importance, attribution has become one of the most misunderstood concepts in modern marketing. It is not flawed, nor has it outlived its usefulness. The difficulty is that marketers increasingly expect it to answer questions that it was never designed to answer.
As paths to conversion have become more fragmented, privacy restrictions have increased and organizations have demanded greater accountability from their marketing investments, the role of attribution has evolved. Understanding how it works remains essential. Understanding where it stops being useful may be even more important.
Why Attribution Became So Important
To understand attribution properly, it is worth remembering why it became such a dominant framework in the first place.
One of digital marketing’s greatest advantages has always been accountability. Unlike traditional channels, where teams often relied on aggregate audience estimates and broad assumptions about effectiveness, digital platforms offered something that appeared revolutionary. For the first time, advertisers could observe individual buying journeys, track clicks, measure conversions and connect activity directly to commercial outcomes. That level of visibility transformed expectations about what marketing measurement could achieve.
As digital channels expanded however, so did the complexity of customer behavior. Consumers no longer discovered brands through a single interaction. They moved between search engines, social platforms, websites, emails, review sites and increasingly multiple devices. What once looked like a straightforward path to purchase became a web of interconnected touchpoints.
As a result, a new question emerged: If multiple marketing interactions contribute to a conversion, which one should get the credit?
The Evolution of Attribution Models
The simplest approach seemed to be last-click attribution. When a conversion occurs, attribution models attempt to determine how much credit should be assigned to the marketing interactions that preceded it. Under the last-click model, all of that value is given to the final interaction before purchase. It became the default for many advertising platforms because it was easy to implement, easy to explain and straightforward to report. Its weakness, however, is equally obvious: every earlier interaction disappears from the story, making it virtually impossible to see how effective your whole marketing arsenal is. A customer may have engaged with multiple campaigns, visited the website several times and conducted extensive research before converting. None of that activity receives recognition.
By contrast. first-click attribution attempted to solve the opposite problem by assigning all credit to the initial interaction. This certainly provided greater visibility into demand generation activity and highlighted channels responsible for introducing customers to a brand. However, it created its own blind spots by disregarding every interaction that followed.
In the intervening years, more sophisticated models have emerged: “Linear Attribution” distributed credit equally across touchpoints. “Time-Decay models” gave greater weight to interactions occurring closer to conversion. “Position-based models” attempted to recognize both the beginning and end of the customer journey. More recently, “machine-learning-driven” approaches have sought to use statistical analysis to estimate the relative importance of different interactions.
Each model offers a different perspective, but none provides a definitive answer. This is not necessarily a weakness of attribution, just a reflection of the fact that customer decision-making is inherently difficult to reduce to a mathematical formula.
Enterprise Attribution Models
As attribution has matured, more advanced statistical models have also emerged. Rather than relying on where a touchpoint appeared within the buying journey, these approaches attempt to estimate how much each interaction genuinely contributes to the final outcome. Two of the best-known examples are Shapley Value and Markov Chain attribution.
The Shapley Value model is borrowed from game theory, an idea that attempts to allocate credit fairly across every marketing interaction. Rather than giving all the credit to the first or last touchpoint, it calculates how much each channel contributes across every possible combination of interactions. The result is a more balanced view of the customer journey, particularly where multiple channels work together to influence a purchase.
Although Shapley Value can provide a more nuanced picture than traditional attribution models, it still relies on observed customer journeys rather than proving cause and effect. It estimates each channel’s contribution based on statistical relationships, not whether a particular interaction genuinely changed customer behavior. For organizations managing complex, multi-channel campaigns, it can be a valuable analytical tool, but it should still be complemented by incrementality testing and broader business measurement.
Similarly, Markov Chains is a statistical model that evaluates how customers move between different marketing touchpoints before converting. Instead of assigning credit based on position within the journey, it measures what happens when an individual channel is removed from the sequence. If conversion rates fall significantly when a channel is excluded, the model assumes that channel made a meaningful contribution to the final outcome.
Markov models often provide a more realistic representation of complex buying journeys because they recognize that channels work together rather than in isolation. Even so, they remain attribution models rather than proof of causation. They estimate the importance of each touchpoint within an observed journey but cannot determine with certainty whether a customer would have converted anyway. As with Shapley Value, their greatest value comes when they are used alongside incrementality testing and Marketing Mix Modeling (MMM) –a statistical approach that estimates how different marketing investments contribute to sales or revenue at a business level, rather than following individual customers– as part of a broader measurement framework.
The Question Attribution Cannot Fully Answer
For many years teams searched for increasingly sophisticated attribution methodologies in the hope that they would provide greater certainty. Yet as attribution models evolved, a more fundamental difficulty began to emerge. Attribution can tell us what happened before a conversion but it can’t say with any certainty what caused the conversion.
At first glance this may sound like a subtle distinction. In practice it is one of the most important concepts in modern marketing measurement. Imagine a customer who first encounters a brand through YouTube, subsequently sees several Meta advertisements, reads organic content, performs multiple searches and finally converts after clicking a branded search advertisement.
An attribution model can identify every touchpoint within that journey. It can distribute credit across those interactions using whatever methodology has been selected. What it cannot reliably determine is which interaction clinched the sale.
The YouTube campaign may have created awareness. Meta may have influenced consideration. Equally, the customer may already have decided to purchase before clicking the branded search advertisement.
The journey can be observed, but the underlying causality is much harder to establish. And this is where many organizations begin to encounter difficulties.
Attribution is often treated as though it measures impact. In reality, it primarily measures participation. Those are not the same thing. Indeed it is just one part of a broader measurement strategy. As discussed in How to Measure Digital Marketing Performance, meaningful reporting connects customer journeys with pipeline influence and commercial outcomes rather than platform metrics alone.
Correlation Versus Causation
The distinction between participation and impact becomes particularly important when evaluating marketing effectiveness. Many channels interact with customers who are already highly likely to convert. Retargeting provides a useful example.
A prospect visits a website, demonstrates clear purchase intent and subsequently sees a retargeting advertisement before converting. Most attribution models will allocate some degree of credit to the retargeting activity because it formed part of the observed journey. But perhaps the more important question is: did the retargeting campaign actually change behavior? And would the customer have converted anyway?
A similar challenge often arises with branded search. Customers who are already seeking a specific company frequently click a paid search advertisement immediately before converting. Attribution models naturally reward the advertisement because it represents the final interaction. Yet the existence of the click does not necessarily prove that the advertisement generated additional demand. This is one of the reasons attribution frequently overstates marketing contribution.
One long-standing criticism from organic search practitioners is that paid search often receives disproportionate credit for the conversion, even though the customer spent 30 minutes researching the product via Google and ChatGPT. Increasingly, attribution also needs to account for the way Paid Media and Organic Search Work Together, because buyers rarely experience channels in isolation.
While attribution is very good at showing which marketing activity was present before a conversion took place, it’s much less reliable at proving which activity actually changed a customer’s mind. Presence and influence are not the same thing, yet they are often treated as though they are. The growing influence of AI Search and Agentic Discovery makes this even more important, as buying journeys become less linear and increasingly difficult to observe through traditional tracking alone.
Enter Incrementality
As marketers became increasingly aware of attribution’s limitations, attention began shifting towards a different question. Rather than asking which channels received credit, organizations started asking whether marketing activity changed behavior at all. This is the domain of incrementality.
Incrementality attempts to measure the additional outcomes created by marketing activity. Instead of examining paths to conversion after they occur, incrementality seeks to understand what would have happened if the marketing activity had never existed in the first place. This is a fundamentally different approach to measurement.
Through methodologies such as holdout testing, geo experiments, audience suppression and conversion lift studies, organizations can begin to isolate the true impact of specific activities. Rather than measuring participation in a journey, they measure behavioral change.
The distinction is significant.
- Attribution tells us who was present.
- Incrementality helps us to understand who made a difference.
However, neither methodology is perfect. Both involve assumptions and limitations but incrementality addresses questions that attribution alone struggles to answer.
Why Marketing Mix Modeling Has Returned
At roughly the same time that incrementality testing gained momentum, another measurement discipline began experiencing a resurgence. Marketing Mix Modeling has existed for decades. Long before digital marketing as we know it today emerged, organizations used econometric models to estimate the contribution of different marketing activities to business outcomes.
Yet for a period, MMM fell out of favor. Digital attribution appeared to offer something superior. Why rely on statistical Modeling when individual customer journeys could be observed directly? The limitations became increasingly apparent as digital ecosystems evolved. Privacy regulations expanded. Third-party cookies began disappearing. “Walled gardens” reduced visibility; buying journeys became increasingly fragmented across platforms and devices.
The more advertisers attempted to measure everything at the user level, the more difficult measurement became. MMM offered a different perspective. Rather than focusing on individual conversions, it examined aggregate business outcomes. Instead of asking which touchpoint received credit, it sought to understand the contribution different marketing investments make to revenue, sales or growth over time.
In many respects, MMM answers strategic questions that attribution was never designed to address. It is less concerned with journeys and more concerned with outcomes.
Attribution, Incrementality & MMM are not in Competition
One of the most common misconceptions in modern measurement is that organizations must choose between attribution, incrementality testing and Marketing Mix Modeling. In reality, each serves a different purpose.
Attribution helps teams understand paths to conversion and the touchpoints involved in conversion. Incrementality helps determine whether activity genuinely changes behavior. MMM provides a broader view of how marketing contributes to business performance at an aggregate level. Seen together, the methodologies are not competing alternatives. They are complementary perspectives.
- Attribution explains participation.
- Incrementality explores causation.
- MMM evaluates contribution.
Together they provide a far richer understanding of performance than any individual methodology can achieve alone.
The Future of Marketing Measurement
For much of the past two decades, organizations searched for the perfect attribution model. Increasingly, it’s becoming clear that no such model exists. Buying journeys are too complex. Human behavior is too nuanced. Privacy constraints are too significant. The modern marketing ecosystem simply cannot be fully explained through a single measurement framework. The future of measurement is therefore unlikely to be dominated by attribution alone.
Instead, organizations are increasingly building broader measurement systems that combine multiple methodologies. Attribution provides directional insight into customer journeys. Incrementality validates causal impact. Marketing Mix Modeling informs strategic investment decisions. First-party data strengthens all three.
The objective is not perfect measurement – that has always been an illusion. Rather, the objective is better decision-making.
Conclusion
Attribution remains one of the most valuable tools available to marketers, but it is only one lens through which performance can be understood. The organizations that measure performance most effectively will combine attribution, incrementality testing and Marketing Mix Modeling, using each where it provides the greatest insight. Together they offer something more valuable than perfect reporting: the confidence to make better commercial decisions.
