The hidden dangers of relying on return on advertising spend as your primary marketing metric

Return on Advertising Spend (ROAS) remains one of the most ubiquitous metrics in the modern digital marketing landscape, serving as a fundamental compass for ecommerce leaders and media buyers alike. By calculating the ratio of sales attributed to specific ad campaigns against the costs associated with those ads, businesses can theoretically map their growth trajectory and optimize budget allocation. However, as the digital ecosystem becomes increasingly fragmented and privacy-centric, the industry is reckoning with a fundamental flaw: ROAS is only as reliable as the attribution model that powers it. Without a nuanced understanding of incrementality, companies risk misallocating millions of dollars toward campaigns that would have yielded results regardless of ad presence.

The mechanics of ROAS are deceptively simple: (Sales Attributed to Ads) divided by (Cost of Ads). While this provides a clear snapshot of performance, it often fails to distinguish between correlation and causation. Mike Murphy, vice president of marketing at the attribution firm Incremental, argues that this oversight constitutes the "Achilles heel" of modern digital advertising. When marketers rely exclusively on last-touch attribution—a model that assigns 100% of the credit to the final advertisement a consumer clicked before purchasing—they inadvertently inflate the perceived success of their campaigns.

The Evolution of Attribution Challenges

Historically, attribution was a relatively straightforward exercise. Before the proliferation of cross-device shopping and privacy-first browser restrictions, a user’s path to purchase was linear and largely traceable. Today, the customer journey is a complex, non-linear web. A consumer might discover a product via a sponsored post on Instagram, research it through a search engine on a mobile device, and finally complete the transaction on a desktop computer.

In this environment, attribution models have struggled to keep pace. When platforms like Amazon or Google claim a 5:1 ROAS, they are often operating within a "walled garden." If a retailer spends $10,000 on retail media and receives $50,000 in reported sales, the initial impulse is to scale that investment. However, if the attribution model ignores organic demand—sales that would have occurred without the ad—the retailer may be paying for exposure that provided no marginal value.

The history of this dilemma dates back to the early 2010s, as programmatic advertising began to dominate. As tracking cookies became less reliable due to browser-level privacy updates from Apple (ITP) and Google (Privacy Sandbox), marketers were forced to rely more heavily on modeled data. This shift sparked a debate that continues today: are these models capturing true consumer behavior, or are they merely optimizing for the platform’s own self-reported success metrics?

The Critical Role of Incrementality

The core issue lies in the distinction between attributed sales and incremental sales. Incrementality refers to the lift in total sales that can be directly attributed to a specific advertising campaign that would not have happened otherwise. If a shopper was already on a retailer’s website with the intent to buy, the ad they clicked may have been a redundant touchpoint.

For example, consider a brand selling electronics on a major marketplace. Data indicates that when these products appear in search results, they often garner organic clicks. If the company launches a retail media campaign, the ads may appear in the same search results. If the attribution system grants the ad credit for every sale occurring after a click, it ignores the fact that a significant portion of those buyers were likely to purchase the product regardless of the advertisement.

Recent industry studies suggest that in competitive retail categories, the "cannibalization" rate—where paid ads displace organic traffic—can reach as high as 40% to 60%. When this factor is ignored, a ROAS that appears to be 5:1 may, in reality, be closer to 2:1 or 3:1 once organic baseline sales are stripped away. This discrepancy represents a massive miscalculation of efficiency and can lead to significant budgetary waste.

The Under-Attribution Paradox

While ROAS is often accused of over-crediting ads, the inverse is equally problematic: under-attribution. Because of increasing privacy restrictions, many consumer touchpoints go dark. If a user views an ad on one device but purchases on another, or if the conversion happens outside the "lookback window" of the tracking pixel, the system fails to log the success.

Google and Meta have responded to this challenge by implementing "conversion modeling," an AI-driven approach designed to estimate conversions that cannot be observed directly. While Google maintains that this modeling is essential for maintaining accurate performance data in a privacy-first world, critics argue that it introduces a layer of abstraction that makes it difficult for brands to verify their true return on investment. If a business relies solely on these modeled figures, they may be underestimating the effectiveness of their advertising, causing them to pull back spending on campaigns that are actually driving profitable growth.

Testing and Validation Strategies

To mitigate these risks, sophisticated marketers are moving toward rigorous testing protocols. For smaller enterprises, the "holdout test" remains the gold standard for validating incrementality. By ceasing advertising for a specific subset of products or a specific geographic region for a set period, brands can establish a baseline of "organic" sales. By comparing this control group against the active campaign, they can isolate the true incremental impact of their spend.

Larger organizations often leverage more advanced methods, such as synthetic control groups or geographic market testing, where advertising is toggled on and off in specific cities. These methods, while resource-intensive, provide a far more accurate picture than standard platform reports. The goal is not necessarily to eliminate ROAS as a metric, but to contextualize it within a broader framework of testing and validation.

The P&L as the Ultimate Source of Truth

In the pursuit of granular, real-time marketing data, many companies have lost sight of the most basic indicator of financial health: the Profit and Loss (P&L) statement. While digital dashboards provide minute-by-minute updates on clicks, impressions, and ROAS, they are often siloed from the broader business reality.

If a company increases its advertising budget by 20% and sees a corresponding 20% rise in top-line revenue, the campaign appears to be a success. However, if that spending increase results in a contraction of net profit margins due to high customer acquisition costs, the campaign is failing at the bottom line. Marketing leaders are increasingly being encouraged to look at "Marketing Efficiency Ratio" (MER)—total revenue divided by total marketing spend—as a broader, more holistic view of performance. Unlike ROAS, which is specific to channels, MER considers the business as a whole.

As Mike Murphy emphasizes, the P&L does not lie. It captures the net result of all efforts, including the costs of goods sold, overhead, and the true cost of acquisition. For modern ecommerce brands, the future of performance measurement lies in balancing the granular insights of digital attribution with the undeniable reality of top-line revenue growth and bottom-line profitability.

Implications for the Future of Marketing

The industry is currently in a transition period. The era of "easy" attribution, characterized by high-fidelity tracking and simple conversion counting, has ended. As platforms continue to gate their data and regulators enforce stricter privacy standards, the reliance on modeled data and probabilistic measurement will only increase.

For brands, the implication is clear: the era of "set it and forget it" advertising is over. Success now requires a more analytical, skeptical approach to platform data. Companies that thrive will be those that treat their ad platform reports as mere suggestions rather than gospel. By fostering a culture of continuous testing, investing in incrementality studies, and keeping a constant eye on the P&L, businesses can navigate the current uncertainty.

The obsession with ROAS as a solitary KPI has led many into a trap of superficial optimization. While the metric remains a useful tool for day-to-day campaign management, it must be supported by a more profound understanding of customer behavior and business outcomes. In the end, the most effective marketers will be those who recognize that while algorithms can track the path of a user, only the business owner can determine the true value of a sale. As the digital advertising landscape continues to evolve, the ability to discern real growth from phantom attribution will be the defining skill of the next generation of commerce leaders.