The Anatomy of the Divide

In 2019, Adidas Global Media Director Simon Peel gave an interview to Marketing Week that has since achieved a kind of canonical status in marketing circles. The company, he explained, had allocated 77 percent of its media budget to digital performance channels and 23 percent to brand. The annual budget in question was approximately two billion euros. The outcome was seven consecutive quarters of declining sales in North America and Europe, loss of market position in China to competitors Anta and Li-Ning, and what Peel described as a systematic error in how the company was measuring its own marketing effectiveness.

"We had a problem that we were focusing on the wrong metrics, the short-term, because we have fiduciary responsibility to shareholders," Peel said. The sentence contains the architecture of the entire problem. The metrics were wrong. The short-term was being optimized. The fiduciary responsibility was real. These facts coexisted with the two billion euros producing worse results than a different allocation might have.

The question the Adidas case poses is not whether the company made a mistake. It is more specific than that: which accounting system made the mistake inevitable?

The standard marketing budget exists in two parallel columns. On one side: performance marketing. This includes paid search, retargeting, programmatic display with direct response creative, affiliate, and conversion-optimized social. Its unit of measurement is the conversion, the click, the ROAS figure that appears in the dashboard within hours of a campaign launching. On the other side: brand marketing. This includes television, high-quality print, out-of-home, sponsorships, and what practitioners sometimes call "upper-funnel digital" — awareness-focused video, content partnerships, influencer programs that optimize for reach rather than clicks.

The problem is not that these two activities require different metrics. It is that the performance column has a real-time accounting system and the brand column has a retroactive one. The performance column shows results as they happen. The brand column shows results only after the brand has accumulated enough equity to manifest as a measurable conversion — which typically takes between eighteen months and three years, by which point the budget decision has long since been made and the attribution has long since been assigned to whatever last click preceded the purchase.

This is the Division. It is not a strategic insight. It is a measurement artifact that has calcified into organizational structure. The performance column appears accountable because it generates numbers. The brand column appears unaccountable because its numbers arrive late, are ambiguous, and can be contested. The result is that every budget cycle, the performance column argues for its own expansion using metrics the brand column cannot produce, and loses.

Les Binet and Peter Field established in their 2013 Institute of Practitioners in Advertising study "The Long and the Short of It" that the optimal budget split — across thirty years of IPA Effectiveness Awards data encompassing roughly one thousand campaigns in eighty categories — is approximately sixty percent brand building and forty percent sales activation. Their follow-up work, "Effectiveness in Context" (2018), refined this and showed that while the ratio shifts by sector (online-heavy markets trend toward fifty-fifty, B2B toward fifty-four-forty-six), performance does not become the dominant allocation in any sector. The forty percent is not brand's share because brands are irrational. It is brand's share because activation without prior brand building produces diminishing returns as the audience that could be activated exhausts itself.

The sixty-forty finding has been replicated, contested, debated, and refined. Byron Sharp at the Ehrenberg-Bass Institute argues the specific ratio overstates what the science can support. But even Sharp, whose work emphasizes reach and physical availability over emotional brand building, does not argue that brand-equity activities are valueless. He argues that their value manifests differently — through the brand's position in the consumer's consideration set at the moment of decision, not through an immediate measurable response. The disagreement is about mechanism, not about whether upper-funnel investment and lower-funnel investment are both necessary.

What is not in dispute is the organizational outcome of having only one measurable column. When only the performance column produces real-time numbers, the budget conversation becomes a conversation between a visible number and an argument. Over time, the argument loses.

What Last-Click Actually Measures

Academic research has documented the systematic bias embedded in last-click attribution with unusual precision. In a 2016 study in the International Journal of Research in Marketing, De Haan, Wiesel, and Pauwels found that last-click attribution underestimates content-integrated activities and yields budget allocations that produce ten to twelve percent less revenue than model-based allocations. Their model-based allocation, using the same underlying data but distributing credit according to the actual contribution of each touchpoint, yielded a twenty-one percent revenue increase.

The mechanism is straightforward: last-click assigns all credit for a conversion to the final ad a consumer clicked before purchasing. This is logically coherent for a retargeting impression shown to someone who has already visited a website. It is logically incoherent for brand-building activity that creates the underlying consideration in the first place. A consumer who sees a Nike television advertisement, searches for Nike the next day, clicks a Google text ad, and purchases — the television advertisement receives zero credit. The Google text ad receives all of it. The attribution system has not malfunctioned. It has done exactly what it was designed to do. The problem is that what it was designed to do is measure a narrow slice of the actual causal chain.

Kireyev, Pauwels, and Gupta, in the same journal issue (2016), found that display advertising exposures increased branded search activity by approximately fourteen and a half percent within a four-week window — despite receiving zero credit in last-click frameworks. They estimated that last-click underestimates display advertising's contribution by approximately thirty-five percent on average. Barajas et al. (2016) found that last-touch attribution undervalues display and native programmatic advertising by eighty-seven percent compared to controlled experiments.

These are not minor calibration errors. They represent a systematic redirection of budget away from activities that are generating substantial value and toward activities that are getting credit for value they did not create.

The performance column, in other words, is not winning the attribution debate because it is generating more value. It is winning because it is better at claiming value generated elsewhere.

The Incrementality Gap

The logical response to this is to run controlled experiments — incrementality tests — that establish what would have happened without the campaign. Lewis and Rao (2015), in a landmark Quarterly Journal of Economics study based on twenty-five large field experiments with major U.S. retailers and two point eight million dollars in ad spend, found something that complicates this solution severely: the median confidence interval on ROI in these experiments is over one hundred percentage points wide. The effects of advertising are small relative to the noise in the data. Detecting reliable ROI, they concluded, often requires more than ten million person-weeks of exposure.

The implication is that incrementality testing — the gold standard for proving causal impact — often cannot produce statistically reliable results at the individual campaign level. It can establish average effects across large budgets and long time horizons. It cannot tell a performance marketing manager whether their specific campaign this quarter produced a genuine increment or captured conversions that would have occurred through organic search.

Gordon et al. (2019), studying fifteen Facebook campaigns with five hundred million user-experiment observations and one point six billion ad impressions, found that observational attribution methods often failed to produce the same effect sizes as randomized experiments. In approximately half of the studies, the estimated percentage increase in purchase outcomes was off by a factor of three across all observational methods.

The Division of Imaginary Budgets is thus not simply a measurement problem with a known solution. It is a structural condition. The brand column cannot produce real-time numbers. The performance column's real-time numbers are systematically overstated. Incrementality testing can correct for this overstatement but often lacks the statistical power to produce definitive results at the campaign level. The budget conversation therefore defaults to what can be measured, not to what is true.

The Adidas Case in the Accounting Frame

What Adidas experienced is predictable from the accounting structure. When a company allocates seventy-seven percent of its budget to performance and twenty-three percent to brand, the performance allocation will appear to justify itself quarter after quarter — because last-click attribution will credit the performance campaigns with conversions that the brand campaigns made possible. The brand campaigns will appear to underperform not because they are ineffective but because their effectiveness is invisible to the measurement system that is being used to evaluate them.

This creates a feedback loop. Performance appears to work. Brand appears not to work. Budget shifts toward performance. Performance now has more inventory to optimize against, so its attributed conversions increase. Brand has less budget, so its reach and frequency decline, and its already-invisible contribution to consideration becomes even more difficult to measure. The Division widens.

Peel's admission about fiduciary responsibility is the crux. The accounting system was producing numbers that satisfied the quarterly reporting requirement. Those numbers were not wrong in the sense of being fabricated. They were wrong in the sense that what they were measuring — the last click before conversion — is not the same thing as marketing effectiveness. But the system had no mechanism for showing this distinction at the time the budget decision was made.

The three-year lag between brand investment and measurable brand equity effect means that the organization cannot reclassify the two billion euros of misallocated spend. The Division does not merely misallocate current budget. It renders past misallocation invisible by the time its consequences become legible.

The Budget That Cannot Be Named

BCG published research in 2025 finding that cutting one dollar of brand spending costs one dollar and ninety-two cents in future investment to regain the lost market share. This is a measurement of what brand equity is worth when it is lost, not when it is built. It implies that the brand column has a real financial value that does not appear in any quarterly report until the moment it begins to disappear.

The GfK and Thinkbox television advertising ROI study found that short-term ROI from television advertising is approximately 1.15 — meaning each dollar spent returns one dollar and fifteen cents in short-term sales. Long-term ROI from the same television advertising is approximately 2.65. The long-term figure is two point three times the short-term figure. Both numbers are real. The budget system is designed to see only the short-term one.

In October 2025, Les Binet presented new IPA effectiveness research arguing that budget size is approximately eight times more important as a driver of profit than ROI optimization. The finding is a direct implication of the Division: if organizations cannot accurately measure ROI, then the organizations that can are actually measuring the wrong thing. The companies that outperform competitors in marketing effectiveness are not those with the best ROAS. They are those with the largest budgets — because large budgets produce the reach and frequency that generate the long-term equity that eventually manifests as conversions, and the conversion is what gets credited to whatever performance activity preceded it.

This is the final structure of the Division. The performance column does not merely claim credit for brand's work. It claims credit in a currency — short-term ROAS — that is genuinely less valuable than the currency the brand column actually produces — long-term market position. The performance column is rewarded for optimizing a metric that is a reliable indicator of marketing effectiveness only when marketing effectiveness is not being systematically misattributed.

The Organization as Measurement System

The practical consequence is that most D2C brands, when they allocate ninety percent of budget to performance marketing, are not making a rational choice based on proven ROI superiority. They are making an irrational choice based on available measurement. The IPA data shows that sixty-forty is the empirically optimal split. The academic literature shows that last-click systematically overcredits performance and undercredits brand by thirty to eighty-seven percent depending on channel. The incrementality literature shows that even rigorous testing often lacks the statistical power to definitively distinguish real performance from captured organic demand.

The organizations that escape the Division share a common characteristic: they treat the gap between what performance marketing claims and what it actually produces as a known distortion, not as a reliable measurement. They use Marketing Mix Modeling to estimate the baseline — what sales would have been without any marketing — and they use that baseline to establish what performance marketing is actually incrementing. They accept that this requires accepting estimates rather than certainties, and they make peace with reporting numbers that their CFO cannot verify in a dashboard in real time.

The organizations that remain trapped in the Division do not have worse performance marketing. They have a measurement system that is structurally incapable of accounting for what brand building does, and an organizational structure that punishes the column that cannot defend itself in the language the system speaks.

The two billion euros Adidas spent is not a cautionary tale about the dangers of performance marketing. It is a cautionary tale about the dangers of a measurement system that can only speak one language and calls the resulting budget allocation a strategic decision.

References