The measurement industry's AI race has a blind spot: the people who actually run the media
From independent attribution dashboards to platform-built models like Meta's Robyn and Google's Meridian, an interpretation layer is getting bolted onto historical data faster than anyone is asking the people who do the work whether that interpretation holds up.
By Chris Marine, founder and CEO, Campfire
Somewhere in product roadmap meetings this year, at more than one measurement company, someone decided that typing a question into a chatbot was the natural next step for an attribution dashboard. Ask it something like "what happens if I move 10% of budget from paid social to connected TV (CTV)," and it answers instantly, in a fluent, confident sentence that reads like a strategic recommendation. This isn't one vendor's idea. It's happening across the independent attribution space and inside the platforms themselves, and I think the industry is moving faster than it's thinking.
I said as much directly to one of our measurement partners this summer, after learning they were building a conversational AI layer on top of their dashboard. I'm not against AI. I'm against an industry racing to build an interpretation layer without building a feedback loop with the people who actually do the work that feeds it: the media planners, the local sales reps, the creative teams, the account leads who know why a plan looks the way it does. The dashboard doesn't call any of them before it answers.
What's being automated, and what isn't
Picture a dashboard being read not just by a marketing team, but by a CFO or president who has said plainly that marketing isn't their area of expertise. An AI-generated answer to "what happens if we shift budget" doesn't arrive with a caveat. It arrives as an answer. That's how a historical data pattern turns into an executive talking point, and how a talking point turns into a decision nobody on the media team actually made.
This isn't limited to independent attribution vendors bolting AI onto their dashboards. The platforms themselves are moving into the same territory. Meta built Robyn, an open-source, AI and machine learning powered marketing mix model, to help advertisers measure cross-channel effectiveness without relying on user-level tracking. Google built Meridian, its own open-source marketing mix model, and in February 2026 added a no-code Scenario Planner that lets marketers model budget shifts without writing a line of code. Both are genuinely useful pieces of open-source work, built and maintained by real research teams, and I don't think either company built them in bad faith.
But when Matt Hertig, CEO of the measurement firm ChannelMix, first looked at Meridian, his read was blunt: Google was rolling out a tool primarily tailored for its own products. The same structural question follows every platform entering this space. A model that's easiest to calibrate using a platform's own reach and query data has a natural gravity toward crediting that platform's own channels. That's not a conspiracy. It's just what happens when the entity building the interpretation layer also owns some of the inventory being interpreted, and it's exactly the kind of thing a feedback loop with outside practitioners would surface early, if anyone built one in.
Attribution can tell you what happened. It can't tell you why.
This is the part of the measurement conversation that gets skipped over most often: attribution and marketing mix models can tell you what happened. Neither can tell you why a media plan was built the way it was.
Neither one knows about the creative that wasn't ready in time. Neither one knows a competitor changed pricing in one region. Neither one knows a channel was deliberately underweighted in a market for three months to test a new format before scaling it. These models see a channel's output. They have no visibility into the conversations, tradeoffs, and business realities that shaped the input, because none of that ever got asked of the people who lived it.
That distinction matters more now, not less, because the tools delivering the output are getting more persuasive. A number sitting in a spreadsheet invites scrutiny. A number delivered in a fluent, conversational sentence invites belief. From where I sit, attribution and modeling should set up better conversations with clients, not replace them. Historical data should help us ask smarter questions. Strategy still has to come from people who understand the full business context, and right now, very few of these interpretation layers are built with a way for those people to push back before the answer reaches a boardroom.
Understanding the full business context is, in a lot of ways, the whole job. Media strategy was never really about picking channels. It's about understanding people and the business dynamics behind a brand's growth. Channels are just where that understanding gets executed.
The place this shows up most clearly: local media
Nowhere is the gap between data and context wider than in how the industry talks about local media, and nowhere is the fix more obvious once you look closely.
Start with what "local" tends to mean when it gets invoked in a national media plan: small. A market to fill in on a spreadsheet. But running a campaign in every market isn't the same as being known in any single one of them. National media often assumes it's already thinking local simply because it's everywhere. It isn't. Local media strategy is about meeting people where they actually are, whether that's a city block or a rural county, and building enough familiarity there that a brand feels like it belongs rather than like it's passing through.
That distinction shows up in the trust data too. Edelman's 2026 Trust Barometer found that trust is consolidating around what's close and familiar: people trust their neighbors (64%), their coworkers, and their own employer's leadership far more than they trust distant institutions. Over the past five years, trust in national government leaders fell 16 points and trust in major news organizations fell 11 points, while trust in neighbors, family, and coworkers each gained ground. Pew Research Center found a similar pattern specifically around news: in a survey fielded in September 2025, 70% of U.S. adults said they trust information from local news organizations, compared with 56% for national outlets, a gap that has held for years even as both numbers have slipped.
None of that is happening in a vacuum. The same Edelman research ties the retreat toward the familiar directly to economic pressure: inflation is now cited by more people (54%) than almost any other single event as a force eroding trust in institutions, and roughly two-thirds of employees say they're worried trade policy and tariffs will hurt their employer. People are tired, and that fatigue is landing on brands as much as on institutions. Distant, generic messaging doesn't land well with a tired, distrustful audience. Familiar, specific, local messaging has a better chance.
The people behind the "traditional" channels
Here's where the obvious interpretation runs out. Yes, people don't consume local newspapers, cable, and radio the way they did twenty years ago. That's been said often enough that it barely counts as insight anymore. What gets left out of that sentence is who's doing the work to keep those channels relevant: the teams inside local broadcasters, radio groups, and publishers who've spent the last several years building out digital and social operations to reach the same communities on new platforms.
BIA Advisory Services projects U.S. local advertising spend will reach $184.5 billion in 2026, and for the first time, digital channels account for more than half of it. The Local Media Consortium's most recent industry survey found that 61.5% of local media companies plan to increase their digital revenue budgets this year, with 72% reporting digital revenue was already up or flat in 2025. That growth doesn't happen by accident. It's the product of local sales and content teams, the same ones who used to sell only spot cable or drive-time radio, now also producing short-form video, managing social accounts, and building first-party audience data for their markets.
Local media was never only the linear channel. It was always the relationship: the local sales rep who knows the market, the content team producing something a national buy never could. Digital and social are just newer rooms in the same house.
Why the supply path matters as much as the channel
This is also where a direct supply path stops being a nice-to-have and starts being the difference between spend that's measurable and spend nobody can fully account for. The Association of National Advertisers found in its most recent programmatic transparency benchmark that $26.8 billion in global media value is lost every year to supply chain inefficiencies, up from $20 billion just two years earlier. Its own recommendation to marketers included shortening the path and establishing direct relationships with the people actually selling the media.
That's precisely what working with a local sales rep does that an exchange buy can't. A direct relationship means knowing exactly which station, which daypart, which show, and which market a dollar bought. It means creative and sponsorship opportunities a programmatic pipe was never built to offer: a live read, a community sponsorship, a co-branded segment. And it means an accountable, traceable line between spend and outcome, which is the actual foundation measurement needs before any dashboard, AI-powered or otherwise, gets involved.
How local media attribution actually gets done
None of this works if measurement stops at "we bought local media and awareness went up." A defensible local attribution approach usually triangulates a few signals rather than leaning on one:
Geo-matched testing, comparing a market running a campaign against a similar market that isn't, to isolate the media's effect from broader trends.
Foot traffic and location data tied to specific stations, dayparts, or flight windows, rather than a campaign as a whole.
Promo code, call tracking, or redemption data that ties back to a specific creative or market, so a station read or a local sponsorship can be measured on its own terms.
Branded search and site visit lift by market, comparing markets on air against markets that are dark.
Coincident timing tests, turning a channel on or off in specific markets to observe the change, rather than assuming a national average applies everywhere.
No single one of those proves the whole effect on its own. Disagreement between them is useful information, not a problem to explain away. That discipline, more signals, more scrutiny of what each one can and can't say, is the opposite of what a chat box answering a budget question in one confident sentence encourages.
The point underneath all of it
I don't think AI belongs nowhere near measurement. I think it belongs in the room after the people who do the work do, not instead of them. An interpretation layer built without a feedback loop to media planners, local sales reps, and creative teams isn't just unhelpful when it gets something wrong. It actively shapes decisions, budget shifts, executive talking points, client confidence, in a direction nobody actually chose. That's the disruption, and it's already underway, one dashboard update at a time.
Media strategy was never just channel selection. It's understanding people and the business dynamics behind a brand's growth, then finding the media environments, local and national, digital and analog, where that understanding can actually do something. Attribution and modeling, AI-assisted or not, are tools for sharpening that understanding. The moment they start standing in for it is a mistake the industry has made before. This version just has better production values, and a much larger feedback gap to close.
