If there was a theme running through this week's advertising news, it wasn't AI. We've all heard enough AI sermons to last several lifetimes. It was execution. The companies actually building the future are pulling away from the ones still issuing press releases about it. Publicis isn't talking about transformation—it's monetizing it. Google is quietly rewriting the rules of measurement while everyone argues over transparency. Hollywood is chasing creators it ignored for a decade. Private equity claims ad tech is too risky one minute, then offers a 50% premium for Criteo the next.

Even the IAB is trying to solve a problem that's plagued this industry since someone first uttered the phrase "digital ecosystem" with a straight face: getting everyone to speak the same language. Meanwhile, agencies continue their evolution from idea factories into consulting firms, and a new women's baseball league demonstrates that sometimes the smartest branding move is simply not overthinking it. The ad industry loves to tell itself it's living through unprecedented disruption. Maybe. But this week looked a lot more like a sorting mechanism. The companies building infrastructure are getting stronger. The companies selling buzzwords are getting louder. Those are not the same thing.

Arthur Sadoun must be wondering if anyone plans to put up a fight. Another quarter, another guidance increase, another pile of blue-chip business, and another reminder that Publicis seems to be operating several years ahead of the rest of the holding company world. The headline isn't just nearly 5% organic growth or wins like Microsoft and American Airlines. It's that 87% of the company's revenue now comes from AI-powered marketing services, making AI less of a shiny demo and more of the engine under the hood. While rivals continue talking about "AI transformation" as though it's a destination they'll someday reach, Publicis has quietly turned it into a business model. The LiveRamp acquisition suddenly looks less like an expensive shopping spree and more like another piece of a larger strategy to own the data, identity, media, and technology stack that clients increasingly need. Publicis is behaving less like an agency holding company and more like an operating system for marketing. That's a much harder business to compete against than another network of creative shops.

Nothing is more entertaining than watching Hollywood congratulate itself for embracing creators roughly a decade after creators stopped needing Hollywood. Suddenly, YouTube libraries are acquisition targets, talent agencies are racing to sign TikTok stars, private equity firms are funding creator IP, and executives are speaking reverently about "Creator Hollywood" as though they invented the concept. They didn't. Creators built audiences while Hollywood was busy rebooting intellectual property it had already rebooted twice. But before everyone starts declaring the death of traditional entertainment, there's an inconvenient reality. Building an audience isn't the same as building enduring entertainment. Most creators thrive because they produce quickly, adapt instantly, and answer only to their communities. Hollywood's superpower remains slowing everything down until every risky idea becomes a committee-approved compromise. There will absolutely be breakout creator franchises, but there will also be a mountain of expensive disappointments greenlit because executives confused subscriber counts with storytelling. The winners won't be the biggest influencers. They'll be the ones who evolve from content creators into genuine world builders.

Advertising has accomplished something remarkable over the last twenty years: it turned buying video into a graduate-level linguistics course. Ask five people to define Connected TV and you'll receive seven different answers, three acronyms you've never heard before, and at least one PowerPoint with circles connected by arrows. The IAB has finally decided enough is enough, rolling out a framework intended to standardize how the industry classifies digital video inventory. It's an admirable goal because AI agents can't automate media buying if the humans feeding them can't agree on what they're buying in the first place. The challenge, of course, isn't creating standards. The advertising industry has always loved standards—provided they don't interfere with anyone's proprietary methodology, unique inventory, differentiated measurement system, or quarterly revenue targets. Every platform believes it's special. Every seller believes its inventory deserves its own category. The IAB deserves credit for trying. Whether the industry follows along is another story entirely.

Marketing mix modeling has returned from the dead, thanks largely to privacy changes that made traditional attribution increasingly unreliable. Google knows this, which is why it's aggressively pushing Meridian while Meta appears to be quietly backing away from Robyn. The pitch sounds irresistible: open-source measurement, greater transparency, and a smarter alternative to the industry's long-standing addiction to last-click attribution. And to be fair, moving beyond click-based measurement is unquestionably progress. But advertisers should remember that "open source" and "independent" are not interchangeable concepts. Meridian still draws much of its value from Google's unique data ecosystem, and many agencies are building measurement products on Google's code without clients even realizing it. None of this is scandalous. It's exactly what a dominant platform should do if it's trying to strengthen its ecosystem. But when the same company owns the media, provides the data, develops the measurement framework, and encourages everyone to adopt it, skepticism isn't cynicism. It's due diligence.

Nothing exposes conventional wisdom faster than someone willing to write a very large check. On one hand, industry observers argue that private equity has grown hesitant to invest in ad tech because AI is moving too quickly for traditional valuation models. Investors know the future belongs to AI-driven marketing infrastructure; they simply don't know how to price companies whose competitive positions can change every six months. Then, almost on cue, Vista Equity Partners and Quinti Capital reportedly offer to buy Criteo at a premium exceeding fifty percent. Apparently ad tech isn't so broken after all. The contradiction is only superficial. Commodity software is becoming less valuable. Infrastructure is becoming dramatically more valuable. Identity, retail media, commerce data, and measurement are increasingly the toll roads of digital advertising, and whoever owns the toll roads usually does just fine. The real risk isn't investing in ad tech. It's investing in yesterday's definition of ad tech.

Hiring Harsh Kapadia to lead a new enterprise creative division isn't really about creative leadership. It's about economics. Agencies have spent years watching consulting firms march into boardrooms, absorb digital transformation budgets, and leave the creative work as an afterthought. Now the agencies are returning the favor. Code and Theory's new enterprise unit is designed to help major clients like Adobe and IBM weave creativity into large-scale transformation initiatives, but the bigger story is what that says about the modern agency business. Campaigns have become the price of admission. Transformation is where the money lives. The distinction between agency, consultancy, systems integrator, technology partner, and business advisor grows blurrier every quarter. Eventually we'll stop pretending they're different businesses altogether.

Perhaps the most refreshing story of the week came from a place few people expected. The Women's Pro Baseball League unveiled four team identities inspired by real women who broke barriers in sports, medicine, and civil rights, and somehow managed to accomplish something increasingly rare in branding: it trusted the audience. Instead of drowning every logo in layers of manufactured symbolism or wrapping every design choice inside a twelve-page purpose manifesto, Arrivals + Departures created brands that feel authentic because they're rooted in authentic stories. Modern branding has developed an unfortunate habit of explaining itself to death. Every color choice becomes a metaphor. Every font becomes a movement. Sometimes the smartest creative decision is simply creating something memorable, meaningful, and wearable. Imagine that.

Stay Bold. Stay Curious. Know More Than You Did Yesterday.

A giant company looks at the wreckage of cookie tracking, the privacy crackdowns, the browser changes, the creeping suspicion across every marketing department on earth that nobody can actually prove anything works anymore, and it does something that looks almost civic. It takes marketing mix modeling, the decades old econometric technique that used to live inside the spreadsheets of a few very expensive consultants, and it gives the code away. Free. On GitHub. Apache license. Audit it yourself. That is Meridian, and that is the pitch.

Google publicly floated Meridian in March 2024 and made it generally available in early 2025, launching alongside a partner program of more than twenty measurement vendors "trained and certified" on the framework. The company's language leans hard on two words: transparent and open. A Google spokesperson, responding to exactly the criticism you are about to read, put it plainly to the trade outlet B&T: Meridian is open-source so anyone can audit the code. The implication, made explicit in the same statement, is that auditability equals independence. You can read the model. Therefore the model is neutral.

Here is the problem with that syllogism. The code was never the conflict of interest. The pipes are.

Let us be fair about what Meridian is, because the technical work is genuinely good. It is a Bayesian model. It handles adstock and saturation curves, the twin facts that advertising keeps working after you stop paying for it and stops working the more you pay. It does geo level hierarchical modeling, which is a real methodological step up from the national aggregates most open MMM tools produce. You can inject your own priors, calibrate against incrementality experiments, and inspect every line of the math. All of that is true and all of that is visible.

None of it tells you where the numbers come from.

An MMM is only as honest as the data you feed it, and Meridian's most valuable inputs arrive pre-loaded from one direction. When you run Meridian through Google's own environment, you get access to what the company calls the MMM Data Platform: core media data for Google channels plus value-add dimensions like Google Query Volume, the indexed search-interest signal Google recommends you drop in as a control variable. The stated reason is sound. Organic brand interest confounds your estimate of what paid search actually did, so you want to model it. The unstated consequence is that the single richest, cleanest, most granular data stream flowing into your "independent" model is one only Google can provide, describing performance on channels only Google can sell you.

That is the shape of the whole thing. Open source the scorecard, keep the pen.

If Meridian had stayed a Python package for data scientists, this would be an academic complaint. It did not stay that. Watch the timeline, because the timeline is the argument.

In February 2026, Google shipped Scenario Planner, a no-code interface so marketers without a statistician on staff can ask budget questions in plain language and get answers back. In May, it announced Meridian GeoX, which folds incrementality experiments directly into the model, and Meridian Studio, an enterprise platform for running MMMs at volume on Google Cloud. Then, at Google Marketing Live on May 20, the load-bearing move: Meridian now lives inside Google Analytics 360, the paid analytics tier that enterprise marketing already runs on. Marketing mix modeling is no longer a specialist deliverable that arrives quarterly as a PDF. It is a tab, sitting next to your campaign reporting, in the interface you never leave.

Arriving in the same breath was Qualified Future Conversions, a Gemini-powered metric that predicts sales up to 180 days out from an ad interaction, using early signals like brand searches. Google's own framing is that only 40 percent of Demand Gen conversions land inside the standard 30-day window, so QFCs exist to credit the campaigns whose payoff you cannot yet see. And Google has been explicit that these predictive signals will eventually feed back into Meridian to refine its accuracy.

Read that last sentence again slowly. The advertising platform generates a proprietary AI prediction about future value. That prediction flows into the measurement model. The measurement model tells you where to move your budget. There is no external checkpoint anywhere in that circuit. Google's ads, Google's analytics, Google's model, and Google's forecast are now sharing signals in a closed loop, and the company is selling the loop as unification. It is unification. That is the concern, not the reassurance.

This is the distinction the whole debate turns on, and it is worth saying in one clean line: being able to inspect and modify code is not the same as being structurally independent from the company that sells the media and controls the data.

Meridian is auditable. Meridian is not independent, and no amount of GitHub access changes that, because independence is not a property of source code. It is a property of who owns the inputs, who sets the defaults, and who profits when the model recommends one channel over another. Google owns the inputs. Google sets the defaults. And Google is the largest single seller of advertising inventory on the planet, which means every budget dollar that Meridian steers toward search or YouTube is a dollar that lands in Google's own account.

The company would say, correctly, that Meridian can ingest TikTok, Pinterest, Snap, Meta, and traditional media, and that nothing stops you from modeling non-Google channels. True. But the turnkey, richly instrumented, lowest-friction version of Meridian is the one that leans into Google's data, Google's signals, and Google's cloud. The path of least resistance and the path that flatters Google's inventory are the same path. You are free to walk a harder one. Most teams under deadline will not.

Here is where it gets quietly cynical. Meridian's repo explicitly exists to let firms "set up and run their own in-house models," and consultancies, agencies, and analytics vendors did exactly that, fast. Walk the market today and you will find MMM products sold as "our platform," "AI-powered mix modeling," proprietary-sounding, premium-priced, and in a meaningful number of cases a Meridian fork running in Google Cloud with Google signals as privileged inputs, wearing a different logo.

There is nothing illegal about building on open source. That is what open source is for. But a brand paying an agency for "independent third-party measurement" deserves to know whether the core modeling logic and the privileged data connections originate from the same company that sells the media being measured. Very few of these offerings foreground that. The word Google tends not to appear in the deck. The independence is in the marketing, not the architecture.

None of this would matter as much if MMMs were bulletproof. They are not, and this is not a Google-specific flaw. It is inherent to the method. When ad spend chases underlying demand, or when channels move together, an MMM can post perfectly respectable fit statistics while badly mis-estimating what any individual channel actually contributed. Independent data scientists have shown, using simulations where the true answer is known in advance, that a model can look healthy in aggregate and still get the per-channel story materially wrong, especially when baseline demand is doing most of the work and the media is just along for the ride. Meridian also models at the channel level, not the campaign level, which practitioners note is where the largest optimization lever actually sits.

Now combine that fragility with the incentive structure. When the same platform supplies the data, promotes the turnkey tool, and profits from the recommendation, the pressure runs one direction: toward confident, actionable budget advice, and away from the boring caveats about collinearity and baseline demand that would tell you to trust the output less. The model's honest uncertainty is bad for adoption. Guess which one the interface is optimized to surface.

The real long-term risk is not any single skewed number. It is standardization. When Google's methodological choices, its calibration approach, its treatment of search, its integration of predictive signals like QFCs, become the default way an entire industry asks and answers the question "what is my marketing worth," those choices stop being one vendor's opinion and start being the water everyone swims in. Alternative frameworks that might weigh non-Google channels differently do not get argued down. They get crowded out, because the Google version is the one that is free, integrated, no-code, and already open in the tab.

Skepticism here is not an accusation of fraud, and I want to be precise about that. Nobody needs to be cooking numbers for this to be a problem. The problem is structural and it survives everyone's good intentions: the most convenient, best-resourced, most authoritative-looking version of "independent" measurement in the market is built on the framework, the data, and the cloud of the company with the largest possible interest in the answer. Open code does not neutralize that. It decorates it.

The house is not cheating. The house rewrote the rules so it would not have to.

The full ADOTAT Scorecards. Vendor comparisons that name winners and losers, with the analysis your planning team needs before Q3 budget meetings.

The takes nobody else will publish. Independent means we don't answer to vendors, agencies, or the trade press establishment.

The archive. Every Scorecard, every analysis, searchable. The reference layer for how this industry actually works.

Read by senior leaders at every major holding company agency, GroupM, Publicis, Omnicom, IPG, Havas, Dentsu, and inside the buying and strategy teams at Netflix, Roku, Pinterest, Disney, NBCUniversal, Spotify, Amazon, Meta, and most of the platforms shaping where ad dollars actually go.

If your competitors are reading ADOTAT+ and you aren't, you're planning with less information than they are.

Expense it.

Founding member rate. Locks in for life.

Also, no ads. Zero ads. No Ads.