75% of Marketers Admit Their Measurement Is Broken: IAB State of Data 2026
75% of Marketers Admit Their Measurement Is Broken
A few months ago I spoke with a marketing director at a consumer goods company in Santiago. She'd spent three years building what she called "our measurement system": Google Analytics, Meta Ads Manager, a half-connected CRM, and a closing spreadsheet that pulled everything together at the end of the quarter.
I asked her how much of her investment she could defend with real incrementality data — not delivery data, not last-click attribution. She went quiet. Then she said: "Maybe 40%. The rest I defend with historical trends and agency pressure."
She wasn't alone. In 2026 the IAB published its State of Data Report, its broadest annual analysis of the state of marketing measurement. The number that resonated across the industry: between 60% and 75% of surveyed marketers admit their systems fail on at least one of three dimensions — coverage, consistency, or confidence. And the harshest data point: not a single respondent said their Marketing Mix Modeling covers all of their paid channels.
Not one.
That number deserves a moment of pause before you keep reading.
3 Reasons the Problem Is Structural, Not About Tools
When a data point like this shows up in an industry report, the first instinct is to look for someone to blame: the agency didn't integrate the data properly, the vendor sold something that didn't work, the internal team didn't have budget. All of those explanations have a grain of truth. But the problem predates all of them.
1. Measurement systems were designed for a world that no longer exists
The measurement infrastructure most companies use in LATAM — and worldwide — was built during the era of third-party cookie tracking. A world where the user journey was, in theory, traceable from start to finish: they saw the ad, clicked, bought, and the system logged the entire path.
That world is over. Third-party cookie deprecation, the rise of ad blockers, iOS cross-app tracking restrictions, and attention fragmenting across dozens of simultaneous channels broke the system's central assumption. But the dashboards are still there, reporting with the same confidence as always, as if nothing had changed.
The result is measurement that measures what it can measure, not what matters. The IAB documented this precisely: the average coverage of current measurement systems falls far short of 100% of spend. The gap isn't a technical problem you fix with a data connector. It's the legacy of systems designed for linear journeys and eternal cookies.
2. MMM exists, but it covers only a fraction of real spend
The IAB report documented something practitioners already knew but rarely said out loud: even companies with active Marketing Mix Modeling fail on coverage. Not a single respondent said their model captures every channel they invest in.
Why? Because building an MMM that integrates broadcast TV, streaming, radio, OOH, digital display, search, social, and e-commerce — with the temporal and geographic granularity needed to be actionable — is a data engineering problem before it's a statistical one. Most models capture digital well because that's where the data is clean, standardized, and API-accessible. Offline gets left out — at best estimated with proxies, at worst invisible.
And if offline is left out of the model, the model systematically underestimates its contribution. The mix tilts toward digital in the next budget cycle. Overinvestment in performance creates a self-sustaining illusion of high ROAS, until organic growth runs out and no one understands why sales stop responding to spend.
This is the cycle the IAB is documenting. It isn't new, but now there's industry data confirming it at scale.
3. Incentives aren't aligned toward integration
There's a third problem, less technical but just as real: almost no one in the ecosystem benefits from measurement being integrated and honest.
Digital platforms — Meta and Google chief among them — have reporting systems that maximize the conversion credit attributed to their own channels. Not out of malice, but by design: each platform measures its own contribution from its own system, with no view of the other channels. The result is that if you add up the ROAS across every platform you invest in, the total usually exceeds 100% of real sales. Each one attributes more to itself than it actually did. If you want to dig deeper into this point, we covered it in detail in What If Your Marketing Metrics Are Lying?.
The LATAM Angle: The Same Problem, Amplified
The IAB's numbers are global, but the problem is amplified in Chile, Mexico, Colombia, and Peru for two concrete reasons.
First, the media mix in LATAM remains deeply multichannel. Broadcast TV and streaming combined account for between 20% and 40% of the mix in consumer categories. Radio maintains frequency and reach at low cost in markets where digital penetration isn't universal. OOH generates between 5% and 15% of incremental sales in impulse categories. All of these channels show up in the investment spreadsheet of any mid-sized brand in the region. Almost none of them show up in a serious performance measurement setup.
Second, marketing teams in LATAM are smaller, and measurement tool budgets are proportionally lower. A company spending 50 million Chilean pesos a month doesn't have the same access to data infrastructure as a global retailer. Without that infrastructure, building an MMM that integrates offline isn't just expensive — it's logistically complex.
What the IAB documents as a global average is, in LATAM, double the problem: the coverage gap is wider, and the resources to close it are scarcer. The practical result is that many brands in the region operate with attribution models that capture, at best, 40% of their real mix.
What a Modern MMM Does in This Context
The obvious response to this diagnosis is to implement MMM. But the IAB is also honest about this: having a model isn't enough if the model only covers half the channels.
An MMM that works in a 2026 context has to meet three conditions that most current models don't meet simultaneously.
Offline data integration. TV, radio, and OOH need to enter the model with their own investment time series, with control variables for seasonality and events. Robyn, Meta's open-source framework, allows for this level of integration. It isn't trivial to implement, but it's possible — and it's what separates a model that says something actionable from one that just reports what you already knew.
Frequent updates. A model that runs once a year isn't operationally useful for deciding next quarter's budget. Modern infrastructure, with TimesFM for time-series forecasting and automated retraining processes, allows for weekly or biweekly update cycles.
Calibration against incrementality tests. A well-built MMM doesn't replace geo tests or incrementality tests — it uses them as priors to calibrate the model's coefficients. The result is a system that improves its accuracy over time, instead of accumulating silent bias.
What an MMM Does NOT Do, No Matter How Well Built
This section matters as much as the previous one, because overestimating what MMM can do is as dangerous as not having it at all.
An MMM doesn't tell you which specific creative to use. It doesn't answer which message resonated best with which audience — that's what creative tests within each platform are for.
An MMM doesn't replace geo tests. Geographic incrementality tests remain the most direct proof of causality. MMM uses those results as calibration input; it doesn't eliminate them.
An MMM doesn't work with less than 6 to 12 months of consistent historical data. If the data has gaps, methodology changes midstream, or periods without investment in some channel, the model will struggle badly to estimate reliable contributions. Starting to measure well today is the prerequisite before the model.
And the most important point: an MMM doesn't replace the team's judgment. The model's coefficients are estimates with confidence intervals, not absolute truths. The decision of how much to invest in each channel still belongs to the team, informed by the model, not delegated to it.
What to Do About This Next Week
The IAB report isn't an invitation to pessimism. It's a diagnosis with two possible readings.
First: if 75% are failing, the gap between those who measure well and those who don't is a real competitive advantage for whoever closes it first. Brands with a better view of their incremental mix make better budget decisions — not in the abstract, but in concrete dollars that stop being spent where they don't generate additional sales.
Second: if no model covers every channel, the goal isn't perfect coverage but relevant coverage. An MMM that integrates 80% of spend with good temporal granularity is infinitely more actionable than an Ads Manager that reports 100% of clicks but zero causality.
The first practical step is an honest inventory: what percentage of your total investment enters your primary measurement system today, not as tracking data, but as a variable in a contribution model that estimates incrementality? Most of the brands we work with at Neuroradar reach that number and are surprised. The average is between 45% and 60%.
That figure, more than any other number in the IAB report, is the real starting point for improving measurement.
If you want to see what that inventory looks like in practice, we have a template in the demo dashboard: dashboard.neuroradar.ai/dashboard-demo. No login, no form. If you'd rather discuss it in the context of your own mix, 25 minutes is enough to surface 2 or 3 hypotheses about where your biggest blind spots are.
Closing
The measurement problem isn't that the tools are bad. It's that we're using tools designed to answer different questions than the ones we need answered today.
A platform dashboard answers "how many conversions were attributed to this channel?" What the CFO asks is "how many additional sales did this investment generate?" Those are different questions. The IAB just documented that 75% of the industry knows it — and that almost no one yet has the answer to the second one.
The good news is that the distance between those two questions has never been shorter than it is now.
Frequently Asked Questions
What did the IAB State of Data 2026 reveal about marketing measurement?
Between 60% and 75% of marketers surveyed by the IAB admit their measurement systems fail on coverage, consistency, or confidence. The most significant data point: not a single respondent said their Marketing Mix Modeling covers all the paid channels they invest in. Source: IAB State of Data 2026.
Why don't measurement systems cover the full mix?
Current systems were designed for linear journeys with third-party cookie tracking. With cookie deprecation, iOS restrictions, and fragmentation across multiple channels, they measure only a fraction of the real mix. Offline — TV, radio, OOH — rarely enters the models even though it can represent between 20% and 40% of the total mix in consumer categories.
How much of the budget can be defended with real incrementality in LATAM?
Based on our experience with brands in LATAM, the average is between 45% and 60% of total spend. The rest is decided using platform history, agency pressure, or intuition — none of which is real incrementality.
How much historical data does an MMM need to work properly?
The recommended minimum is 6 to 12 months of consistent historical data with the same temporal granularity (weekly or daily) across all channels. With less data or gaps in the history, the model's confidence intervals are too wide for reliable investment decisions.
What's the difference between what a platform reports and real incrementality?
A platform reports conversions attributed to its own channel using its own attribution model. Incrementality measures how many additional sales an investment generated compared to a scenario without that investment. If you add up the ROAS reported by all your platforms simultaneously, the total usually exceeds 100% of your real sales — each platform attributes more to itself than it actually did.