The great metrics deception: ROAS, AI and real profit
Why numbers are lying and how to return to real profit
Increasingly sophisticated dashboards, reports full of upward-trending charts, and platforms promising continuous optimization can give the impression of marketing under control. Today, however, this representation increasingly coexists with a very different reality: shrinking margins, rising operating costs, and decisions based on numbers that tell only part of the story.
The gap between what metrics show and what financial statements reveal comes from how data is selected, interpreted, and used to guide business decisions.
Dashboards can lie. Financial statements do not.
Scenario and context
In today’s digital landscape, marketing lives with a clear paradox: platforms promise millisecond-level optimization, dashboards have never been richer, and charts almost always appear to point upward. Yet behind this façade, many companies are seeing their margins shrink. At HT&T Consulting we analyzed financial and marketing data from more than 50 e-commerce and lead generation companies over the past 18 months. The pattern is consistent: growing budgets, stable or increasing ROAS, and an average net profit decline of 22%.
Dashboards can lie. Financial statements do not.
Data is not lacking. What is missing is the meaning given to it. If today you make decisions primarily based on ROAS, last click, and impressions, you are often not doing strategic marketing: you are optimizing metrics that no longer have a direct connection to real profit.
ROAS: from guiding metric to false friend
For years, ROAS was the absolute reference point. Today, in most cases, it has become a vanity metric—not because it is technically wrong, but because it is incomplete and often distorted.
ROAS measures how much attributed revenue advertising platforms report. It does not measure whether the company is actually making money.
ROAS tells a convenient story. Margin tells the truth.
Platform-calculated ROAS embeds a structural conflict: the platform has a vested interest in proving that spend generates value. What ROAS does not see is everything that happens after and around the sale: logistics costs, returns, packaging, customer care, category-level margins, and organic cannibalization.
Margins are not lost in campaigns. They are lost in the sum of costs campaigns ignore.
The most critical issue is incrementality. Many campaigns perform well simply because they intercept already-mature demand, taxing a conversion that would have happened anyway.
A campaign is incremental only if it generates sales that would not have occurred without it: attribution is not incrementality.
Concrete example: A premium cosmetics brand celebrated a ROAS above 6 on retargeting campaigns. By analyzing raw data, more than 40% of users already had the product in their cart before clicking the ad. When campaigns were paused, revenue dropped only marginally, while net margin increased significantly.
If you turn off a campaign and do not lose sales, you were not growing—you were paying a tax.
The myth of last click in a multi-touchpoint world
Today the purchase journey is no longer linear and, above all, no longer attributable to a single touchpoint. A user may discover a product through social content, explore it via a generative search engine like ChatGPT or Gemini, discuss it within a community or with colleagues, and only days later complete the purchase through a newsletter or a branded search.
In this scenario, assigning conversion value to the last click is a convenient but deeply misleading simplification. Last click does not measure the real contribution of a channel to the decision-making process. It measures only the final event before conversion.
This reading creates an illusion of efficiency: channels that intercept already-formed demand appear to “perform better,” while those that build awareness, trust, and purchase intent seem expensive and hard to justify in reports.
Optimizing for last click means investing where risk is lower, not where value is created.
The outcome is almost always the same. Companies over-invest in the bottom funnel, where conversion is more likely and measurable, and under-invest in demand generation. In the short term, this strategy appears effective. In the medium term, however, the pool of new users shrinks, audiences become saturated, and acquisition costs rise.
When the funnel is no longer fueled from the top, marketing stops creating growth and starts simply competing for the same people—more expensive, harder to convert. That is when last click stops being an analytical shortcut and becomes a strategic limitation.
Impressions and vanity metrics: paying for noise
Visibility has become abundant. Real attention, however, is increasingly scarce: the web is saturated with content, partly generated artificially, that produces impressions but not trust.
Impressions are noise. Authority is signal.
We have observed cases where AI-based response systems cited competitors with minimal advertising investment simply because they were more authoritative and semantically coherent. In this context, millions of impressions do not move the income statement and often distract from the metrics that actually matter. The observatory we launched is the clearest reflection of this new reality. Taking the automotive observatory as an example, new Chinese brands that are not yet advertising in Italy are overtaking established brands spending millions across multiple channels.
The new hierarchy of metrics: returning to margin
Survival belongs to those who manage to bring marketing back into the income statement. The starting point is no longer platform-attributed revenue, but the real contribution margin generated by each marketing activity.
This shift in perspective is not only methodological, but cultural. It means stopping the evaluation of marketing as a cost center to be optimized and starting to read it as a financial lever that must generate sustainable value over time.
POAS (Profit Over Ad Spend). POAS answers a simple and often avoided question: are advertising campaigns generating profit or just revenue? Unlike ROAS, POAS incorporates real operating costs and forces a confrontation with actual margins.
POAS separates campaigns that grow from those that consume margin.
When POAS falls below a sustainability threshold, apparent growth quickly turns into erosion. In such cases, increasing budget does not amplify results—it accelerates losses.
MER (Marketing Efficiency Ratio). MER offers a holistic and less manipulable view of performance. It measures whether marketing is working for the company as a whole, not for a single platform or channel.
If MER falls while spend rises, growth is artificial.
A declining MER signals growing dependence on ads and demand that cannot sustain itself. In other words, marketing is buying short-term results without building structural value.
nCAC (New Customer Acquisition Cost). nCAC measures how much it costs to acquire a customer who has never purchased before. It is a critical metric for scalability because it separates real growth from simple recycling of existing customers.
Scalability begins when nCAC is under control.
Confusing the acquisition cost of new customers with that of returning customers is one of the most expensive mistakes in modern marketing. If nCAC exceeds the margin of the first sale, retention is not an option—it is a requirement. And without solid retention, no growth model can hold over time.
AI Share of Voice: the new invisible competition
A growing share of purchase decisions now flows through AI assistants and response engines. The challenge is no longer appearing first, but being recommended.
AI Share of Voice measures how often your brand is cited as a relevant solution by an artificial intelligence system compared to competitors.
It is not who ranks first on Google that wins, but who is recommended by AI.
If AI does not recognize you as an authoritative source, for a growing portion of the market you simply do not exist—regardless of how much you invest in advertising.
From visibility to recommendation: a matter of method
AI Share of Voice is not a metric to be “checked” once a month. It is the result of a series of structural choices involving content, brand authority, semantic coherence, and the data made available to artificial intelligence systems.
For this reason, at HT&T we do not treat AI Share of Voice as an isolated output, but as a measurable consequence of a working framework that integrates data analysis, marketing mix modeling, content, and source governance.
This is the same approach we use in our vertical observatories and in budget allocation projects based on margin and real incrementality, where the goal is not to appear more, but to be cited and recommended in the contexts that truly matter.
AEO and GEO: when optimization is no longer just for search engines
What we are observing is not simply a channel shift. The way people arrive at purchase decisions is changing. A growing share of these decisions no longer originates from traditional search, but from answers generated by artificial intelligence systems.
In this context, talking only about SEO is no longer sufficient. Two complementary disciplines come into play: Answer Engine Optimization (AEO) and Generative Engine Optimization (GEO).
AEO concerns a brand’s ability to be understood, cited, and recommended by response systems. GEO concerns how generative models interpret content, sources, and authority signals to construct an answer.
If AI cannot explain you, it cannot recommend you.
Optimizing for AEO and GEO means stopping the pursuit of abstract visibility and starting to build content, data, and signals that make the brand readable, reliable, and recommendable in decision-making contexts that truly count.
Case study: when revenue falls and profit grows
A food sector scale-up was investing over €100,000 per month in advertising. High ROAS, negative bottom line: POAS analysis showed that bulky best sellers were structurally unprofitable, while niche products generated high margins.
Reducing revenue is not a defeat if margin increases.
By reallocating the budget, total revenue decreased, but net profit grew dramatically within a few months. This is not a trick—it is the natural consequence of measuring marketing as a financial lever, not as a platform dashboard. And this was achieved using MMM (marketing mix modeling).
Marketing is a financial discipline
Marketing is no longer a creative discipline supported by data. It is a financial discipline expressed through creativity, technology, and content. The future CMO must sit alongside the CFO and speak the same language: margins, cash flow, and long-term value.
Marketing often fails not because it spends too much, but because it measures poorly.
When profit becomes the center, choices change: you distinguish what is incremental from what is merely attributed, set a sustainable ceiling for nCAC, and invest in real authority—the kind that makes a brand citable and recommendable even by AI systems.
Conclusion
Traditional metrics are not wrong. They are simply insufficient. Effective marketing is not the one that generates more revenue, but the one that generates sustainable value.
Growth without margin is not growth. It is erosion.
The right question to ask is: “how much is left in the company after paying everything, including platforms, logistics, and suppliers?” Those who look at these numbers build the future. The others celebrate green dashboards while margin disappears.
Frequently asked questions
Are traditional metrics like ROAS and impressions wrong?
No, they are not wrong in a technical sense. However, they are insufficient when used alone. ROAS and impressions describe what happens on platforms, not what happens in the company’s income statement.
Why can a high ROAS coexist with declining margins?
Because ROAS does not account for operating costs such as logistics, returns, packaging, customer care, and organic cannibalization. Attributed revenue can rise while real margin falls.
What does it mean to measure marketing based on margin?
It means evaluating marketing activities based on the real economic contribution they generate, not only on platform-attributed revenue.
What is the difference between attribution and incrementality?
Attribution assigns a conversion to a channel. Incrementality measures whether that conversion would have happened without that channel. They are different concepts and often confused.
Why is last click a misleading metric?
Because it assigns all value to the final touchpoint before conversion, ignoring the touchpoints that build demand, trust, and purchase intent.
What happens when companies invest only in the bottom funnel?
In the short term, performance may appear to improve. In the medium term, the pool of new users shrinks, audiences saturate, and acquisition costs rise.
What are vanity metrics?
They are metrics that create a perception of success, such as impressions or reach, but have no direct connection to economic value creation.
What is POAS and why is it more useful than ROAS?
POAS measures profit generated relative to advertising spend. Unlike ROAS, it incorporates real costs and allows evaluation of campaign sustainability.
What is AI Share of Voice?
It is a metric that measures how often a brand is cited or recommended by artificial intelligence systems compared to competitors.
Why does brand reputation now also depend on AI?
Because a growing share of purchase decisions is mediated by AI assistants and response engines that interpret and describe brands before users even visit a website.
Further reading and references
-
A clear introduction to Marketing Mix Modeling and its use in allocating budget based on real business impact.
-
A critical analysis of why many traditional marketing metrics fail to represent real economic value.
-
A reference article on the growing gap between operational metrics and strategic decision-making at the management level.
Harvard Business Review – Why Most Marketing Metrics Are Broken
-
An in-depth look at the shift from attribution metrics to measurement models based on causality and real incrementality.
-
A practical explanation of incrementality applied to growth and marketing investment decisions.
-
HT&T’s approach to marketing budget allocation based on margin, real incrementality, and MMM models.
-
HT&T observatory on the evolution of brand visibility within AI-based response systems in the automotive sector.
Continua a leggere
12 minutes of reading
5 minutes of reading
6 minutes of reading
And it consumes less energy.
To return to the page you were visiting, simply click or scroll.