As an Independent Strategic Consultant, Ivan Guzenko provides professional advice to the CEO and C-level management for SmartyAds Inc.

getty
For two decades, no matter what happened in a campaign, there was always an attribution report to explain it: last click, first click or some U-shaped compromise in between. It felt scientific, but now the alibi is running out of evidence to point to.
Mediaocean’s 2026 Advertising Outlook Report, based on surveys of more than 6,100 marketers, showed GenAI has ranked as the most important consumer trend, cited by 70% of marketers. With this, only 19% are using AI for head-to-toe campaign management, versus 43% who use it for data analysis. Marketers trust AI to analyze what already happened, but not yet to decide what should happen next. That gap is where attribution is running out of road and where prediction is starting to take over.
The Stakes Are Rising, The Signals Are Fading
Previously, attribution was built for a simpler customer journey—one browser, one device and one cookie quietly following along. That journey doesn’t exist anymore, as digital identity reached its breaking point after privacy laws have chipped away at the signal attribution depended on.
Sure, the long third-party cookies saga has ended positively for advertisers in Chrome in 2026. Still, other browsers, including Safari, Firefox and Brave, keep blocking cookies by default, affecting a big part of global internet traffic. Attribution models that lean on cross-site tracking are losing signal quality every day, whether or not any single browser flips a switch. The result? Fewer tracking mechanisms mean fewer marketers can figure out how people interact with their ads.
Still, around 42% of marketers (linked in the Mediaocean research above) say that getting good quality data (or being able to access it) is a major obstacle, and only 10% have all their ad technology systems fully integrated. The attribution issue isn’t that marketers don’t care about it anymore, but because the data they need to make it work is getting harder to collect.
The Channel Mix Pressures For Urgency
The Mediaocean research also points to many advertisers planning to spend more money on certain types of media in 2026, like TV and video, social media and AI-driven media. The IAB thinks that spending on TV will increase by almost 14% by the end of 2026. This giant melting pot of data sources becomes hard to track and understand, especially across devices.
To make the most of this money, companies need to be able to measure how well their media is working across all channels. They should be able to view and combine data from many different sources in near-real time. This way, companies can better understand how their media is performing and make good decisions about where to spend their money.
Modeling Without Watching
If tracking every individual’s data isn’t reliable anymore, the industry’s answer shouldn’t be to fix tracking, but to stop depending on it. That’s the logic behind the comeback of marketing mix modeling (MMM): instead of following individual users, it works from aggregate spend and outcome data, sidestepping the privacy problem entirely rather than trying to patch it.
Industry minds think this way of doing things is going to keep growing. Nearly half (46.9%) of U.S. brand and agency marketers plan to invest more in marketing mix modeling over the next year. Moreover, the whole marketing mix modeling market is expected to get almost three times bigger by the mid-2030s.
While this method is good at the big picture (channel-level, often quarterly), the other method, called incrementality testing, will tell advertisers whether one specific tactic inside this or that channel is actually driving results. More than half of U.S. brand and agency marketers (52%) already run incrementality tests, and another 36% plan to invest more in them.
Overall, AI-weighted models can now sit alongside first-touch, last-touch and time-decay approaches, comparing them against the same dataset instead of forcing marketers to bet on one and hope. Meanwhile, data clean rooms are now giving brands an opportunity to merge the data from ad platforms with CRM and analytics data. All this became possible without the raw data ever changing hands, with some vendors claiming accuracy gains of up to 40% from AI models trained inside them.
What Media Buyers Can Do Now
The shift from attribution to prediction requires more than new measurement tools; it also calls for changes in how media buyers prepare their data, evaluate models and act on their insights.
Stop auditing campaigns and start auditing your data pipes.
Predictive modeling isn’t a plug-in you install overnight. It’s a discipline that rewards teams preparing the ground first. Before evaluating any predictive or AI-orchestration vendor, map how cleanly data actually moves between your DSP, your CRM and your measurement layer.
Run MMM and incrementality tests at the same time (not one after another).
This way, you can use MMM to plan your budget with the finance team every quarter and look at how each channel is doing.
Build a human-in-the-loop checkpoint into orchestration, not around it.
The goal isn’t removing marketers from the loop; it’s making sure someone is positioned to catch a model that’s wrong. AI can amplify bad inputs just as efficiently as it amplifies good ones.
Negotiate for explainability.
Ask vendors how their predictive models weight channels and what data feeds them. A forecast you can’t look into is just attribution’s black box wearing a new label.
Re-score your channel mix against fragility.
Prioritize channels with durable data (first-party or contextual data) as they can provide more consistent and trustworthy results.
The Bottom Line
Attribution isn’t going obsolete overnight, but its role is changing from a detailed spreadsheet explaining every slight detail to just one input feeding a model that blends marketing mix modeling, incrementality testing and a human still watching the wheel. The marketers who treat this as an infrastructure rebuild—unifying their data pipes, pairing prediction with a holdout test and keeping someone close enough to catch a confidently wrong forecast—will be the ones best positioned to make more informed media investment decisions as traditional attribution becomes less reliable.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?

