Marketing Leadership minus the hype 2: The measurement trap

Posted by Mike O'Brien on May 26, 2026 in Marketing Leadership, Postcards from the digital frontier | Leave a comment

Why the attribution crisis is a leadership problem disguised as a data problem

The marketing industry’s response to the erosion of third-party cookie tracking has been, in its broad shape, technically impressive and strategically insufficient.

First-party data infrastructure has been rebuilt at considerable expense. Data clean rooms have been deployed to enable privacy-compliant measurement across platforms. Probabilistic modelling has been developed to fill the gaps left by deterministic tracking. New attribution frameworks have been constructed to approximate what the old ones measured with less precision and more acknowledged uncertainty. The industry has, in short, invested heavily in building more sophisticated versions of the system that broke, on the assumption that the system itself was sound and the problem was the data infrastructure supporting it.

That assumption is the central problem this piece wants to examine. And the problem with the assumption is that it is wrong in a way that matters enormously for where the investment goes next.

The attribution crisis is not a data crisis. It is a conceptual crisis: a fundamental misunderstanding of what attribution models were measuring, what they were missing, and what marketing organisations actually need to know to make better decisions. Better data will not solve it. A different measurement philosophy might.

The system that broke was already broken

The uncomfortable argument that the measurement conversation has been largely avoiding is this: the erosion of third-party tracking has not destroyed a reliable measurement system. It has exposed one that was always far more fragile than the industry found it convenient to acknowledge.

It is worth being precise about the technical reality. Google reversed its plan to deprecate third-party cookies in Chrome in 2024 and confirmed in 2025 that it would not introduce a user choice prompt. Third-party cookies remain active in Chrome by default. Safari and Firefox continue to block them, and regulatory pressure on data collection is intensifying across both the UK and US markets. The practical effect is that the measurement crisis is real and accelerating regardless of what any single browser decides, because the combination of privacy regulation, walled garden opacity, and the structural limitations of attribution logic has undermined the system’s credibility independently of any single technical event.

That matters because it means the solution is not primarily technical. The attribution models that defined a decade of digital marketing measurement, last-click, last-touch, and the more sophisticated multi-touch variants that followed, were measuring the visible endpoint of decisions that were made earlier, for reasons the model could not see. They attributed commercial value to the channel that happened to be present at the moment of conversion rather than to the channels, experiences, and accumulated impressions that actually produced the decision to act.

Consider the customer who converts through a paid search click after fourteen months of email newsletter engagement, four visits to a content hub, a positive review read on a third-party platform, and a recommendation from a trusted colleague who had their own experience with the brand. In most attribution systems that customer is recorded as a paid search conversion. The newsletter, the content, the review, the word of mouth: invisible. The paid search click: credited in full, used to justify the paid search budget, and reported upward as evidence of channel performance.

This was never accurate measurement of how marketing value is created. It was measurement of what was technically measurable, dressed in the language of causation and presented with a confidence the underlying methodology could not support. The industry accepted it because it was legible, because platforms reported it with impressive specificity, because it justified existing spend allocations, and because the alternative, acknowledging that the most important things marketing does are largely invisible to the measurement infrastructure, was politically inconvenient in every direction.

The privacy and tracking changes have not created a measurement problem. They have made an existing one impossible to ignore.

Why the wrong answer felt so right

To understand why the attribution model persisted for as long as it did despite its fundamental limitations, it helps to understand what it was actually optimised for, and it was not optimised for measuring marketing value accurately.

It was optimised for producing numbers that could be reported in the format that internal governance structures required. A CMO presenting to a CFO needs to demonstrate return on investment in terms that connect marketing spend to commercial outcomes within a time horizon the organisation’s reporting cycle can accommodate. Attribution models produced exactly that: a direct line, however fictitious, between spend and outcome, reported with apparent precision, within the quarter.

The measurement that would have produced a more accurate picture, tracking the slow accumulation of brand trust, the compounding effect of consistent communications over years, the reduction in customer acquisition cost that follows sustained investment in genuine customer relationships, does not fit that format. It requires a longer time horizon than quarterly reporting accommodates, a tolerance for uncertainty that most governance structures resist, and a willingness to credit investment made in previous periods for outcomes appearing in the current one. None of those things are comfortable in a room where the CFO is asking what last quarter’s campaign generated.

The attribution model did not survive because it was accurate. It survived because it was compatible with the political and reporting structures it served. That is a leadership problem before it is a measurement one.

The extraction model connection

The measurement crisis does not sit in isolation. It is a direct expression of the same management logic that the first piece in this series identified as the underlying cause of the marketing industry’s human cost: the systematic prioritisation of short-term measurable output over the long-term development of value that compounds.

Attribution-based measurement is extraction logic applied to the measurement function. Value only what can be seen immediately, credit only what can be directly connected to a conversion event within the reporting window, and systematically ignore the long-horizon accumulation of trust, relationships, and genuine brand standing that produces the conditions in which any individual campaign either succeeds or fails.

The consequence for marketing leadership is direct and underexamined. Leaders who have built their entire commercial case on attribution metrics have made themselves dependent on measures that can be disrupted by platform decisions, privacy regulation, and technical infrastructure changes that are entirely outside their control. When the measurement breaks, the marketing leader carries the accountability for a framework they inherited and a dependency they did not create. The management logic that produced the dependency is rarely the thing that gets examined.

This dependency is also not confined to measurement. The fourth piece in this series examines how the same logic operates across the broader search and discovery landscape, where brands have progressively outsourced their customer relationships to platform systems with their own commercial objectives. The measurement dependency examined here and the platform dependency examined there are expressions of the same underlying problem: organisations that have allowed critical commercial infrastructure to be controlled by entities whose interests are not aligned with their own.

What measurement is actually for

Before proposing alternatives to attribution, it is worth establishing something more fundamental: what measurement is for.

The dominant answer, implicit in how most organisations have built their measurement infrastructure, is that measurement is primarily about proving marketing happened, demonstrating that spend produced return within a reporting cycle in terms the governance structure can evaluate. That is a legitimate function. It is not the most important one.

Measurement’s more valuable function is improving the quality of marketing decisions, providing the understanding of how value is created, where it accumulates, and what conditions make any marketing investment more or less effective, that allows organisations to allocate better over time. Those two functions look similar from the outside. They produce radically different measurement priorities.

Measurement designed to prove marketing happened will consistently produce metrics that justify existing spend allocations. The channels most invested in are also the channels most measured, and the measurement infrastructure built around them reflects and reinforces the investment logic that created it. Paid search attribution justifies paid search budgets. Social media engagement metrics justify social media budgets. The system is self-referential in a way that makes genuine reallocation toward long-horizon investments extraordinarily difficult.

Measurement designed to improve decision quality will periodically produce uncomfortable findings: that some of the most expensive channels are generating less genuine value than the attribution model suggests, and that some of the hardest-to-measure investments, in brand trust, in direct customer relationships, in genuine content authority, are generating more value than the organisation has been crediting them for. Those findings are politically difficult. They are also the ones most worth acting on.

The alternatives: an honest assessment

There is no single replacement for attribution that solves the measurement problem cleanly. The organisations claiming otherwise are selling something. What exists is a portfolio of approaches, each with genuine strengths and honest limitations, that together produce a more accurate picture of how marketing value is created than any single model can provide.

Marketing mix modelling works at an aggregate level without requiring individual user tracking, making it both privacy-compatible and platform-independent, a significant advantage in an environment where both are under pressure. It measures the contribution of different marketing activities to overall business outcomes over time, capturing some of the long-horizon effects that attribution models miss entirely. Its honest limitations are significant: it requires substantial data over meaningful time periods, it is expensive to do rigorously, and its outputs are directional rather than precise. That last point deserves reframing. Directional insight built on honest methodology is strategically superior to precise reporting built on measurement that was always more fragile than it appeared. The CFO who understands that distinction is a more valuable partner than one who demands precision the data cannot support.

Incrementality testing asks a fundamentally different question from both attribution modelling and mix modelling. Rather than asking which channels were present when conversions happened, or what historical spend correlates with, it asks what would have happened if a specific marketing activity had not run at all. A brand divides its audience or geography into two groups as similar to each other as possible. One group receives the marketing activity being evaluated. The other does not. The difference in outcomes between the two groups measures the genuine causal contribution of that activity, which is the question attribution models never actually answer despite appearing to do so. The results are frequently surprising and politically inconvenient, which is precisely why incrementality testing remains underused relative to its methodological value. Organisations that have run it seriously often find that the channels their attribution model credits most heavily produce less genuine lift than expected, and that investments the model struggles to credit produce more.

Brand tracking connected to business outcomes measures the accumulation of trust, familiarity, and consideration that precedes conversion rather than the conversion itself. When connected explicitly to downstream commercial metrics, customer acquisition cost over time, pricing power, market share in consideration sets, it provides the CFO conversation with something the attribution model never could: a credible account of how brand investment produces commercial return over the horizons that matter. The challenge is building that connection rigorously rather than asserting it, which requires longitudinal data and the patience to let the relationship reveal itself over time.

Qualitative and observational research provides access to the genuine reasons people make decisions, the actual causal account that quantitative models can approximate but never fully reach. Why did this customer choose this brand over the alternatives that were equally available? What role did the content, the community, the recommendation, the remembered campaign from three years ago play in a decision that the attribution model recorded as a direct response conversion? These questions cannot be answered by any tracking system. They can be answered, partially and provisionally, by studying the people who made the decisions, and the organisations that invest in doing so consistently develop a quality of customer understanding that compounds into marketing effectiveness in ways their attribution dashboards will never show.

Most organisations moving toward more robust measurement are combining these approaches rather than seeking a single replacement: mix modelling for strategic budget decisions and directional long-horizon understanding, incrementality testing for genuine causal measurement of specific activities, brand tracking for accumulation measurement, and qualitative research for the why behind the what and when. Each covers limitations the others have. Together they produce a more honest and more useful picture than any single approach.

The invisibility problem

The most durable observation the measurement piece can make is also the most difficult to act on: the most valuable things marketing does are the hardest to see in the data.

Trust accumulated through years of consistent, honest communication reduces customer acquisition cost in ways that show up gradually and diffusely rather than in any individual channel report. Brand familiarity built through sustained presence makes sales conversations shorter and conversion rates higher across every channel, but the attribution goes to the channel at the bottom of the funnel rather than the investment that created the conditions for the conversation to happen. Epistemic authority in a category produces inbound interest from exactly the audiences an organisation most wants to reach, but that interest arrives through channels and pathways that attribution models credit to search, to referral, to direct, rather than to the years of genuine intellectual investment that made the organisation worth seeking out.

Community attachment that generates genuine word of mouth is perhaps the most extreme example. The recommendation from a trusted colleague that actually tips a purchasing decision is completely invisible to every measurement system the recommending brand operates. It cost nothing to generate in the period it was generated. It was the product of an experience that happened years earlier, of trust accumulated through repeated coherent interactions, of the kind of authentic standing that the next piece in this series examines in detail. No attribution model sees it. No dashboard captures it. And the organisation that has invested in producing it cannot point to a single metric that proves the investment worked.

These are not unmeasurable things. They are measurable with different approaches, longer time horizons, and a tolerance for the acknowledged uncertainty that rigorous measurement of complex phenomena always involves. What they are is invisible to the measurement infrastructure most organisations have built, which means that following that infrastructure’s logic consistently produces systematic underinvestment in the activities most likely to generate durable commercial value. Understanding what those activities are, and why they produce the outcomes they produce, is precisely what the authenticity argument that follows this piece is built around.

The CFO conversation

Everything in this piece ultimately comes down to a conversation that happens in a room where the marketing leader is making the case for investment to someone whose primary responsibility is financial stewardship. That conversation is where measurement philosophy becomes leadership practice, and where most CMOs are currently under-equipped.

The CMO who can only speak in attribution terms in that conversation has already lost the argument for long-horizon marketing investment, because the attribution model will never produce evidence that justifies it. Every experiment the system runs is designed to see short-horizon return, so every result it produces confirms the value of short-horizon thinking. The budget follows the evidence, the evidence follows the measurement, and the measurement was designed for a different objective than the one the CMO is actually trying to serve.

The conversation that needs to happen instead is about investment philosophy, and specifically about the difference between measuring the return on the last pound spent and understanding the conditions that make any pound worth spending. Those are different questions, and they require different answers. The CFO who has been given only the first question has a rational basis for preferring the channels that produce the most legible immediate return. The CFO who has been given the second question has a basis for evaluating the long-horizon case: for brand trust that reduces acquisition cost over time, for direct customer relationships that reduce platform dependency and the cost volatility that comes with it, for epistemic authority that shortens sales cycles in ways no attribution model captures.

The specific argument worth making in that conversation has three components. First, the risk argument: precise measurement of the wrong thing is more dangerous than approximate measurement of the right thing, because it produces confident misallocation rather than acknowledged uncertainty. An organisation that knows its measurement is imprecise will apply judgement. An organisation that believes its measurement is precise will follow it, even when it is pointing in the wrong direction. Second, the compounding return argument: brand value, trust, and direct customer relationships accumulate in ways that attribution models cannot see but that eventually show up in pricing power, customer lifetime value, market share, and crisis resilience, and the organisations that have invested in building them are consistently more durable commercial performers than those that have not. Third, the honest acknowledgment of what the proposed measurement philosophy cannot yet do, because the CFO who is told that a new approach solves every problem will be appropriately sceptical, and the CMO who can say “here is what we can measure, here is what we can infer, and here is what we are genuinely uncertain about” is making a more credible case than one who is claiming a precision the new approach does not provide.

Building the internal conditions for that conversation is a leadership priority before it is a measurement one. The CFO who understands what marketing is for, not proving activity happened but improving the quality of decisions about where to invest, is a fundamentally different partner than the CFO who has only ever been given attribution reports. Creating that understanding is the CMO’s responsibility, and it begins long before the budget conversation. It begins in the everyday language used to describe what marketing is building and why, in the metrics chosen for internal reporting, in the way customer relationships are valued on the balance sheet of leadership attention, and in the consistent articulation of an investment philosophy that treats brand value as something real and compounding rather than a soft qualifier to the hard numbers.

Four questions worth asking

The measurement philosophy audit. Is your current measurement infrastructure designed to prove marketing happened or to improve marketing decisions? Map the primary metrics used in internal reporting and ask honestly which objective each one serves. The ratio tells you more about your strategic exposure than any analytics review.

The invisibility audit. What are the three most valuable things your marketing organisation does that your current measurement infrastructure cannot see? Name them specifically, not categories like “brand building” but actual activities: the newsletter that has been running for four years, the content hub that produces the majority of inbound sales conversations, the community that generates the word of mouth your attribution model records as direct traffic. Then ask what it would take to measure their contribution with sufficient rigour to make an internal investment case.

The CFO conversation test. Can you currently make the case for your measurement investment in terms of decision quality rather than reporting accuracy? If you were asked to justify your measurement infrastructure not by the numbers it produces but by the quality of the decisions it enables, what would you say? If the answer is unclear, the internal political conditions for long-horizon marketing investment do not yet exist and building them is a leadership priority.

The false precision question. Where in your current measurement framework are you accepting precision that you know is unreliable because acknowledged uncertainty would be politically difficult? Identifying those places specifically is the first step toward measurement that improves decisions rather than defends them, and the beginning of the CFO conversation the investment requires.

The measurement philosophy the moment requires

The measurement crisis will not be resolved by better data infrastructure. It will be resolved, partially, provisionally, and with honest acknowledgment of remaining uncertainty, by organisations willing to ask a more fundamental question than how to measure what they used to measure with less data.

The question worth asking is what marketing organisations actually need to know to make better decisions over the time horizons that produce durable commercial value. The answer points toward a portfolio of measurement approaches that together illuminate how value accumulates rather than simply where conversions appear. It points toward internal conversations about investment philosophy that treat brand trust, direct relationships, and genuine authority as balance sheet assets rather than soft qualifiers. And it points toward the kind of CFO relationship that is built on shared understanding of what marketing is building rather than shared consumption of attribution reports that both parties know are telling an incomplete story.

That is not a comfortable transition for organisations that have built governance structures, reporting cadences, and internal credibility around the measurement model that is failing. But it is a necessary one, and the organisations that make it seriously will find themselves with a clearer understanding of where their marketing investment is actually generating value, a more credible internal case for the long-horizon investments that compound, and a measurement philosophy that survives the next disruption rather than depending on the technical infrastructure of the last one.

Further reading

Les Binet and Peter Field, The Long and the Short of It – ipa.co.uk | Nielsen on marketing mix modelling – nielsen.com | WARC on effectiveness measurement – warc.com | Analytic Partners on ROI measurement – analyticpartners.com | Byron Sharp, How Brands Grow | Ehrenberg-Bass Institute on brand metrics – marketingscience.info | Harvard Business Review on marketing measurement – hbr.org | Marketing Week on CMO and CFO relationships – marketingweek.com