Nearly six in ten marketers say their organizations are probably or definitely putting money into the wrong places, holding back spend from channels they believe would perform well because they cannot measure them adequately, according to a report AppsFlyer published with EMARKETER on July 16, 2026.

The finding reframes a problem the programmatic advertising industry has treated as a reporting inconvenience into something with a direct cost. When measurement fails, budgets follow the data that exists rather than the data that matters, and the report argues that gap is widening precisely as artificial intelligence takes over more of the decisions about where those budgets go.

The topline number is 58.6%. That share of surveyed marketers agreed, either definitely at 12.1% or probably at 46.5%, that their organization is underinvesting in channels it cannot adequately measure, leaving customers and revenue on the table for competitors. A further 10.2% conceded they lacked the visibility to know whether the problem applied to them. Only 7.6% said the mismeasurement was definitely not happening.

What the survey covered

The research rests on responses from 157 US agency and brand leaders surveyed in May 2026. EMARKETER developed and fielded the study in collaboration with AppsFlyer, publishing it under the title "Mobile-Grade Measurement for the Modern Marketer." The sample is modest, and the survey was commissioned by a company that sells measurement infrastructure, a point worth holding in view when reading conclusions that favor better measurement.

Within those limits, the questionnaire probed three connected areas: how confident marketers feel about attribution on each channel, which channels create the worst blind spots, and what stands between organizations and effective use of AI in their campaigns. The answers describe an industry in transition, aware of what it needs and frustrated by what it cannot yet see.

Confidence is thin across every channel

Asked to rate their confidence in attribution accuracy on a scale from one to five, marketers produced weighted averages clustered in a narrow band. Connected television, or CTV, scored lowest at 3.1. Desktop and web advertising scored highest at 3.6. No channel emerged as a clear winner, and no channel inspired the kind of confidence that would loosen a budget.

The scarcity of top marks tells the sharper story. "Very confident" scores ranged from 11.5% for programmatic display to 17.8% for both social media and desktop or web advertising. At the other end, CTV drew the most "not at all confident" ratings at 10.8%, with commerce media close behind at 10.2%. Most marketers landed in the tentative middle, expressing some confidence but not enough to commit incremental dollars.

When the survey asked which channels create the most significant blind spots in cross-channel attribution, only 7.0% of respondents chose none of the above. The channels named as blind spots by the most marketers were social media at 50.3% and CTV or streaming video at 47.1%. Both are among the fastest-growing destinations for ad budgets, which sharpens the contradiction: marketers are pouring money into environments they trust least to measure.

That tension has surfaced repeatedly in reporting on the streaming channel. Recent survey work found that CTV wins budget but not trust, with a majority of buyers doubting platform-reported performance claims even as investment climbs. Separate research documented that IP-based CTV targeting fails roughly three times in four, a structural accuracy problem that ripples through frequency management and attribution well beyond the television screen.

Mobile as the reference standard

The report's organizing argument is that mobile advertising already solved a version of this problem, and that its methods now transfer to other channels. The survey defined mobile-grade measurement as measurement capabilities built to handle privacy constraints, fraud, and cross-device identity. On that definition, 93.0% of brand and agency leaders recognized the need for such measurement or reported already implementing it across other channels.

The breakdown behind that figure shows work in progress rather than completion. Nearly a quarter, 24.2%, are actively building these capabilities now. Another 23.6% have applied the approaches partially, and 14.7% have done so for most or all channels. An overwhelming 93.0% majority said they planned to bring a similar playbook to non-mobile channels or had already begun.

Ran Avrahamy, chief marketing officer at AppsFlyer, grounded the claim in the technical history of the channel. "Mobile apps operated under constraints that no other digital channel faced at the same intensity," he said. "The rising investment in this channel is proof that it's working."

The constraints he describes are specific. Mobile measurement partners had to deliver accurate numbers and detect fraud while navigating different privacy frameworks between Android and iOS, a complexity that intensified after Apple introduced its AppTrackingTransparency policy in 2021. Connecting a click or view on one platform, a download through an app store, and a first open in a separate environment with no persistent cookie demanded a matching methodology that web attribution never required. "No web analytics tool faces that structural split," Avrahamy said.

The response, according to Avrahamy, was to run parallel methodologies, including deterministic matching, probabilistic modeling, and referrer-based attribution, and to decide in real time which signal to trust and at what confidence threshold. "The result was a measurement infrastructure that had to stay accurate under signal loss, validate data quality continuously, and model truth from incomplete information," he said.

The report offers spend growth as external validation. According to EMARKETER forecasts cited in the study, US mobile app install ad spend rose 13.3% year over year in 2024, with 9.2% growth projected for both 2025 and 2026. In-app ad spend climbed 16.9% in 2024, with 12.7% and 12.0% growth forecast for the following two years. "The sustained growth in mobile app ad spend is the market's confirmation that the infrastructure built to meet those challenges actually works," Avrahamy said.

Yory Wurmser, principal analyst for advertising, media, and technology at EMARKETER, added a demand-side reason the mobile signal runs deep. "Because mobile apps provide consumers with a better experience, consumers tend to engage more consistently with them," he said. "When brands get consumers to download their app, they get more conversions and richer data from their interactions."

The AI amplification problem

The report's most consequential thread connects measurement quality to AI performance, and it is here that the survey moves from familiar territory into the stakes that make the timing matter. Ad workflows are being rebuilt around AI tools that optimize and automate decisions, and those tools inherit whatever flaws sit in the data beneath them.

Asked which data foundation challenge most limits their ability to use AI effectively, 36.3% of marketers named siloed data that cannot be connected across channels, the single most cited barrier. Inconsistent measurement definitions and taxonomies followed at 34.4%. A quarter, 26.1%, pointed to a lack of clean, standardized first-party data, and 24.2% cited identity resolution gaps across devices and platforms. Respondents could select their top two.

Avrahamy framed the risk in blunt terms. "AI doesn't evaluate the quality of its inputs, it only amplifies them," he said. "A model trained on clean, validated, fraud-free signals produces better predictions, sharper audience segmentation, and the right budget allocation. When data is compromised, incomplete, or wrong, the model produces confident wrong answers at scale."

The speed of automated systems is what turns a data problem into an operational one. "If the data foundation isn't solid, mistakes propagate faster and at greater scale before anyone catches them," Avrahamy said. "The feedback loop that would normally surface a bad decision gets compressed or bypassed entirely. The practical consequence is that data governance, which used to be a competitive differentiator, is becoming a prerequisite."

This diagnosis echoes a wider industry conversation. Reporting on how AI can amplify broken measurement rather than fix it has documented the risk that partial data, fed into predictive models, automates the old garbage-in problem at unprecedented scale. Measurement vendors have moved to address the input side directly; one recent product tackles the data-quality problem in marketing mix modeling that many teams did not realize they had, validating campaign files before models are built.

Humans still hold the loop

Despite rapid AI adoption, most organizations are not yet handing over control. Nearly half of respondents, 49.7%, execute AI optimization recommendations without manual review less than half the time, while 27.4% do so rarely or never. That reluctance, the report suggests, marks organizations still mid-transition rather than fully automated.

Avrahamy located the reason for keeping a person in the loop in the nature of what an agent does. "Human judgment involves knowing when to stop and question the premise," he said. "An agent acts, adjusting bids, reallocating budget, suppressing audiences, and the human sees the outcome, not the decision. Bad data doesn't produce inaccurate reports. It instructs agents to take the wrong actions, at speed and scale, with limited human visibility until the damage is done."

The barriers to extracting value from AI reinforce the pattern. Poor data quality from fragmented, incomplete, or unreliable signals was the top-cited obstacle at 22.3%. Explainability, meaning an inability to validate what the AI was actually doing, came second at 17.8%. Tool maturity at 16.6%, talent at 14.0%, and budget constraints at 10.2% rounded out a list that reads as a transformation still underway. Only 4.5% said they were already extracting full value from AI.

The picture is not uniformly negative. In most cases, the AI tools themselves are not seen as the problem: 41.4% of respondents said they were somewhat more confident in overall marketing measurement as AI adoption increased, and another 19.1% said they were significantly more confident that AI had made measurement clearer.

Pressure to prove returns

Running underneath every finding is the demand to justify spend. Half of organizations, 49.7%, are under significant pressure to demonstrate clear ROI across all channels, and 29.9% face extreme pressure. Only 3.8% reported limited pressure. The squeeze is nearly universal, and it collides directly with the measurement gaps the survey documents, because a channel that cannot be measured cannot be defended in a budget review.

The report cites external forecasts pointing to money waiting on the sidelines. Underrepresented channels could see a 5.6% average increase in spend over the next one to two years, according to the Interactive Advertising Bureau. Angelina Eng, the IAB's vice president for its measurement center and center of excellence operations, described the underlying obstacle during a recent IAB webinar cited in the report. "AI is only as good as the data that it receives," she said. "We have lots and lots of data. But as an industry, we do not speak the same language. We are not structuring our data sets the same way."

Wurmser tied the difficulty to the shape of the modern media landscape. "Privacy regulations and a fragmented media ecosystem have made measurement more complex," he said. "Even when you can identify a user, it's often hard to track them as they pass through a variety of touchpoints leading to an ultimate purchase."

Why this matters for the marketing community

The report lands amid a broader shift in how the industry thinks about the relationship between measurement and money. A separate study found that most marketers waste at least 11% of media budgets on optimization signals that are not tied to verified purchase data, with a majority doubting platform-reported results and only a fifth running fully automated optimization. The AppsFlyer and EMARKETER survey supplies the mirror image of that waste: not just money spent badly, but money never spent at all in channels that might have performed.

For agencies and brands, the practical stakes concentrate where measurement and AI intersect. As automated systems assume more of the buying and optimization layer, the quality of the signals feeding them becomes the constraint on the whole system. AppsFlyer's own recent history reflects the strategic weight the industry now assigns to that layer: four of the largest companies in digital advertising invested in AppsFlyer to keep its measurement neutral in a June 2026 deal valuing the company at 2.7 billion dollars, a transaction structured specifically to prevent any single investor from shaping attribution logic.

The channels most in need of the fix are the ones absorbing the most budget. CTV and streaming, named as a blind spot by 47.1% of marketers here, continues to attract performance dollars faster than its measurement infrastructure can support, a gap documented across the sector. Whether the mobile-grade playbook the report advocates can close that gap at scale remains an open question, and one the survey's commissioning by a mobile measurement vendor does not settle. What the data does establish is that the cost of not closing it is no longer hypothetical. It is showing up in budgets that never reach the channels most likely to work.

Timeline

  • 2021 - Apple introduces its AppTrackingTransparency policy, intensifying the privacy complexity mobile measurement partners had to navigate across Android and iOS
  • 2024 - US mobile app install ad spend rises 13.3% year over year; in-app ad spend increases 16.9%, per EMARKETER forecasts
  • May 2026 - EMARKETER fields the survey of 157 US agency and brand leaders in collaboration with AppsFlyer
  • May 21, 2026 - Research finds most marketers waste at least 11% of media budgets on optimization signals not tied to verified purchase data
  • June 22, 2026 - Google, Meta, Moloco, and Unity invest in AppsFlyer at a 2.7 billion dollar valuation in a deal structured to keep measurement neutral
  • July 16, 2026 - AppsFlyer and EMARKETER publish "Mobile-Grade Measurement for the Modern Marketer"

Summary

Who: AppsFlyer, the mobile measurement and attribution company, in partnership with research firm EMARKETER, surveying 157 US agency and brand leaders. Ran Avrahamy, AppsFlyer chief marketing officer, and Yory Wurmser, EMARKETER principal analyst, provided commentary.

What: A report finding that 58.6% of marketers believe their organizations underinvest in channels they cannot adequately measure, tying measurement gaps to budget misallocation as AI takes over more campaign optimization. Social media at 50.3% and CTV at 47.1% were named the worst attribution blind spots.

When: The survey was fielded in May 2026, and the report, "Mobile-Grade Measurement for the Modern Marketer," was published on July 16, 2026.

Where: The survey covered US agency and brand leaders. AppsFlyer is based in San Francisco; EMARKETER is a division of Axel Springer.

Why: As AI systems assume more control over budget decisions, the quality of the measurement signals feeding them determines outcomes. Marketers under significant or extreme ROI pressure withhold spend from channels they cannot prove work, leaving potential revenue unclaimed while automated tools risk amplifying flawed data at scale.