In the modern digital marketing landscape, few metrics carry as much weight—or cause as much internal debate—as Return on Advertising Spend (ROAS). For decades, brands have relied on this straightforward calculation to justify budgets, optimize campaigns, and prove the tangible value of marketing investments to the C-suite.
However, as digital privacy regulations tighten, consumer behavior fragments across multiple devices, and retail media networks (RMNs) dominate the advertising ecosystem, a growing chorus of data scientists and marketing executives is raising a red flag. ROAS, they argue, is only as good as the attribution model powering it. Without accurate attribution, chasing a high ROAS can lead brands to scale failing strategies while starving profitable ones.
Main Facts: The Illusion of Accuracy in ROAS
At its core, ROAS is designed to be a fundamental barometer of efficiency. It connects marketing investment directly to revenue generation through a deceptively simple formula:
$$textROAS = fractextSales Attributed to AdstextCost of Ads$$
For instance, if an ecommerce brand spends $10,000 on a digital campaign and attributes $50,000 in generated sales to that initiative, the resulting ROAS is 5:1. For every single dollar invested, the business ostensibly receives five dollars back.
To the untrained eye—or an executive demanding immediate proof of performance—this appears to be a clear green light to increase ad spend. However, according to industry experts like Mike Murphy, Vice President of Marketing at attribution firm Incremental, this metric harbors a dangerous "Achilles’ heel."
The core issue lies in the limitations of standard attribution models, particularly last-touch attribution. ROAS is inherently misleading when a tracking mechanism gives an ad too much credit, too little credit, or fails entirely to account for baseline purchase intent and incremental lift.
Chronology: The Evolution and Breakdown of Digital Attribution
To understand how modern marketing arrived at this attribution crisis, it helps to examine how tracking has evolved over the past two decades:
- The Early Web Era (Late 1990s–2000s): Digital advertising relied primarily on simple "last-click" models. Whichever banner ad or search result a user clicked immediately before making a purchase received 100% of the credit for the conversion.
- The Multi-Device Boom (2010s): As consumers transitioned from desktop computers to smartphones, tablets, and smart devices, the linear path to purchase shattered. A user might view an ad on Instagram via mobile, research the product on a laptop, and eventually purchase via a tablet. Basic attribution models struggled to connect these dots, leading to fragmented performance data.
- The Privacy Crackdown (2018–Present): Regulatory shifts such as the European Union’s GDPR, the California Consumer Privacy Act (CCPA), and Apple’s App Tracking Transparency (ATT) framework stripped away third-party cookies and identifier cross-app tracking. Platforms were forced to rely on modeled data, and traditional ROAS metrics grew increasingly opaque.
- The Retail Media Explosion (2020s–Present): Retail giants like Amazon, Walmart, and Target built multi-billion-dollar advertising networks. Because these platforms sit directly at the point of purchase where buying intent is exceptionally high, they often claim high ROAS figures—frequently masking whether the ads actually drove new sales or simply captured demand that already existed.
Supporting Data: Overcounting, Undercounting, and Incrementality
The fragility of ROAS becomes glaringly apparent when examined through the lens of incrementality—the measurement of whether an ad campaign drove a net-new sale that would not have occurred organically without the ad’s intervention.
The Overcounting Trap
Consider the earlier example of the ecommerce brand spending $10,000 on retail media with a reported 5:1 ROAS using last-touch attribution.
When analyzing the campaign deeper, the marketing team discovers a critical operational overlap: about half the time the retail media ad appears on a search results page, the brand’s organic listing is already visible in a dominant position.
In these instances, the last-touch attribution model assigns full credit for the sale to the ad, even though the consumer was actively searching for the brand and likely would have purchased organically. When adjusted to account for true incrementality—crediting only the revenue genuinely produced by the ad’s presence—the true ROAS drops from 5:1 down to 3:1.
The Undercounting Blind Spot
Conversely, the attribution problem frequently works in reverse, presenting an artificially low ROAS because the tracking system fails to capture the full customer journey.
Retail media environments again serve as a prime testing ground for this flaw. A shopper might encounter a sponsored product ad on the Amazon marketplace while browsing, but ultimately leave the platform and navigate directly to the merchant’s proprietary direct-to-consumer (DTC) website days later to complete the transaction. Alternatively, the user might transition seamlessly across operating systems and devices, blinding standard platform pixels.
In these scenarios, the ad successfully inspired the purchase intent, but the attribution ecosystem fails to detect it, resulting in an artificially depressed ROAS. Major platforms attempt to combat this with tools like Google Ads’ "conversion modeling," which estimates unobservable conversions based on historical user behavior and privacy-compliant data samples. Without such modeling, reported performance represents only a fraction of reality.
Official Responses and Expert Perspectives
Industry leaders emphasize that marketers must look past self-reported platform metrics and rigorously test the validity of their attribution data.
Mike Murphy underscores the dangers of relying exclusively on surface-level metrics provided by advertising dashboards, noting that standard attribution tools often take credit for sales that were destined to happen anyway.
"ROAS can take credit for organic sales that would have happened anyway, or for a sale an earlier touchpoint drove," Murphy explains. "It also only sees what can be tracked directly."
To counteract these blind spots, experts advocate for controlled experimentation to ground-truth platform data. For brands working with limited ad budgets, Murphy suggests executing straightforward holdout tests. By pausing ad spend on a specific category or group of products for several weeks while maintaining baseline spend on another comparable group, marketers can measure the actual delta in sales performance.
While such tests provide directional insight rather than microscopic precision, they offer a vital reality check against automated dashboard figures. Larger advertisers with deep enterprise resources can leverage randomized or geographic split-test methodologies provided by sophisticated marketing mix modeling (MMM) partners.
Implications: The Bottom-Line Reality and the P&L as the Ultimate Source of Truth
The ultimate implication of the ROAS reliability crisis is a necessary shift in executive mindset. While return on ad spend remains a vital tactical metric for day-to-day campaign optimization, it can no longer be treated as an infallible business ledger.
When digital marketing budgets scale upward, overall gross profit and net revenue should ideally follow a corresponding upward trajectory. Conversely, if ad spend is deliberately pulled back, aggregate sales performance should reflect that reduction.
When platform-reported ROAS and actual business growth diverge, marketers must look beyond specialized attribution dashboards and defer to the ultimate financial arbiter.
As Murphy bluntly summarizes:
"Your P&L should be your first source of truth — it doesn’t lie."
Ultimately, modern digital marketing success requires treating ROAS not as an absolute gospel, but as a hypothesis that must be constantly challenged through rigorous incrementality testing, cross-device modeling, and a relentless focus on bottom-line profitability.
