The Question Most Advertisers Are Afraid to Ask

Earlier this year, a large Mexican retailer did something most marketing teams would consider reckless. It switched off its display campaigns across several states of the country, deliberately, and for nearly two months.

There was no budget crisis behind the decision. The company wanted an answer that no dashboard could give it: how much of their sales actually depend on this advertising?

After more than 15 years in digital advertising, I can count on one hand the advertisers I have seen ask that question with real money on the table. Not because the question is exotic, because both possible answers are uncomfortable. If the answer is “less than we thought,” someone has to explain years of reported returns. If the answer is “more than we thought,” someone has to explain why the budget was ever treated as discretionary. Either way, the comfortable ambiguity disappears. And comfortable ambiguity, I have come to believe, is the most expensive product our industry sells.

What the Metrics Said, and What They Didn’t

This retailer’s campaigns were measured the way most campaigns in Mexico are measured: clicks, click-through rates, sessions, ROAS, cost per acquisition. None of these metrics is the problem. They are essential for optimization, and any team that ignores them is flying blind.

But they all answer the same kind of question: how did each sale happen? Which click, which channel, which path. Attribution distributes credit among touchpoints. What it cannot do, by design, is tell you what would have happened in a world where the campaign never existed. It describes the mechanics of a sale without proving the cause of it.

This is not a Mexican blind spot; it is a global one. The IAB and BWG’s Global State of Data 2026 study found that roughly 3 out of 4 marketers say their measurement systems lack the speed, accuracy, or trust they need. We have more data than at any point in the history of this industry, and less confidence in what it means.

The Experiment

So the retailer ran the only test that answers the causal question: it removed the variable. Display campaigns were paused in a set of test regions while the rest of the country continued as usual. An independent third party measured the outcome using a geo-based synthetic control design, comparing what actually happened in the paused regions against a statistical model of what should have happened.

The result: total e-commerce sales in the test regions fell 19.23% below the expected scenario. In under two months, the gap amounted to tens of millions of pesos.

There are two ways to read that number. The obvious one is that the advertising was working, it was protecting nearly a fifth of online revenue in those regions. The more interesting reading is that no attributed metric had ever expressed that dependency. ROAS had always said the campaign was efficient. It had never said what portion of the business would evaporate without it. Those are different pieces of information, and only one of them belongs in a board conversation.

Attribution Is Not Incrementality

The distinction sounds academic until it costs money. Attribution exists to optimize campaigns: it splits the credit for each conversion across channels and touchpoints so teams can shift budget week to week. Incrementality exists to understand the business: it measures which sales would stop existing if the investment stopped. One is the campaign’s accounting; the other is the business’s audit.

Both matter. Confusing them is where the damage happens. Treat attributed revenue as caused revenue and you will overvalue channels that harvest demand rather than create it. Dismiss advertising because attribution looks weak and you may cut spend that is quietly holding up your baseline, which is precisely what this retailer would have risked doing on instinct alone.

The industry is moving, slowly, toward this discipline. A 2025 EMARKETER and TransUnion survey found that over a third of marketers plan to increase their investment in incrementality testing over the next 12 months, and the major platforms have been lowering the cost of entry for running experiments. Controlled testing is migrating from econometrics teams at global brands to something a mid-sized advertiser in Guadalajara can realistically run. The designs scale with the business: full geographic experiments for large retailers, audience holdouts for retargeting and CRM programs, and simple on-off tests for smaller operations, less precise, but an honest first signal. That shift will not stay optional for long.

What Happened After the Answer

Here is the part of the story that rarely gets told. Once the study confirmed the campaign generated incremental sales, the question changed. It was no longer ,“Does this work?” but, “What, specifically, makes it work?”

That second question is where incrementality stops being an audit and becomes a design input. A display campaign that genuinely moves a business is built in layers: prospecting that reaches people who do not know the brand but resemble its customers, that is where future demand is built; retargeting that recovers intent that already exists, where the quality and safety of the environment matters more than the volume of impressions; and the customer base itself, the most underestimated layer, where the goal shifts from the first conversion to lifetime value. Knowing the campaign’s causal contribution lets you allocate across those layers according to what each one builds, instead of according to whichever one the last click happens to flatter.

Why This Matters for Mexico Now

Digital advertising investment in Mexico has grown faster than the discipline used to evaluate it. E-commerce keeps expanding, budgets keep following it, and boards are asking sharper questions than the dashboards were built to answer. At the same time, every advertiser now has access to essentially the same platforms, the same inventory and the same artificial intelligence. Technology is no longer where the advantage lives.

The next competitive advantage in Mexican digital advertising, I would argue, is epistemological: the discipline to know, not estimate, not narrate, but know what your investment causes. That discipline is uncomfortable, occasionally humbling, and it is available to any company willing to design a serious test.

So here is the question I would leave with any executive reading this. If you switched off your campaigns tomorrow, do you know how much of your sales would fall?

If the answer is no, you are not measuring your marketing. You are narrating it. The retailer in this story spent two months finding out the truth, and the most expensive part of the experiment was not the paused media. It was accepting that what the company had previously called certainty had, all along, been an assumption.