There is a tell in this industry that never lies. When a company is winning, it talks about its clients. When a company is losing, it talks about its architecture.

Spend a month talking to operators across the four big holding companies, which is what ADOTAT just did, and that tell separates the field faster than any earnings release. Publicis talks about the accounts it just won. Omnicom talks about the platform it just relaunched. WPP talks about the technology it is building. Dentsu talks about Japan. Only one of those is a sentence about winning business. The other three are sentences about why the business will come eventually, or why the business that walked out the door was never really the point, or why the business that stayed is the business that always mattered.

ADOTAT spent the past month on this. More than twenty conversations across all four holding companies. Some at the C-level. Some inside the data and identity teams. Some inside the agencies. The most candid conversations came from inside the data infrastructure businesses, where the distance between the press release and the operational reality is the widest, and where one executive said something about a rival's stack so blunt that we are saving it for Part Two because it would end a working relationship if it ran next to their title. None of them were authorized to talk to us. None of them are named in this piece. But that's ADOTAT for you.

The picture they painted is not the picture the trade press has been printing. The trade press has a four-horse race with one front-runner. The operators describe something closer to one company winning and three companies managing their decline at different speeds.

Start with the only objective scoreboard that matters, which is who actually won the business.

Per COMvergence's Final 2025 Global New Business Barometer, Publicis Media generated $10 billion in net new client billings in 2025. That is roughly one-third of all the media spend that changed agencies all year. The nearest competitor, Mediabrands, came second at roughly $1.75 billion. Dentsu third at roughly $1.61 billion. Havas fourth. Omnicom Media Group fifth at roughly $1.32 billion.

Read those numbers again. The four holding companies chasing Publicis are separated from each other by less than half a billion dollars. Publicis is separated from the entire pack by roughly six times. This is not a leaderboard. A leaderboard implies the field is playing the same game. This is one company playing a different game and four companies playing each other for second.

And then there is WPP, which finished the year on the wrong side of zero. The only Big Six holding company with a negative net result. The scale of that number, and the retention figure underneath it, is the single most damning statistic in the entire competitive cycle, and it is the one the trade press keeps wrapping in the word "turnaround." We will get to exactly how bad it is. It is worse than "turnaround" allows.

Let us be precise about WPP, because the charitable framing is doing a lot of work.

The financials are not a soft patch. Revenue down 8.1 percent reported. Like-for-like down 5.4 percent. Reported operating profit down 71 percent after more than £750 million in combined goodwill and property impairments. A 62 percent dividend cut. An exit from the FTSE 100 in December. Market cap below £3 billion, down from a £25 billion peak. The accounts WPP lost in 2025 are not small accounts. Mars. Coca-Cola North America. Paramount. PayPal. Every one of them moved toward Publicis.

Cindy Rose, eight months into the CEO job, has been genuinely honest about the diagnosis. She has named the disease as organizational complexity, the absence of an integrated operating model, and inconsistent execution. She has a plan, Elevate28, with phases labeled "Stabilise 2026, Build 2027, Accelerate 2028." She has notched real recent wins: JLR, Estée Lauder, Mastercard, SC Johnson, Kenvue creative.

Here is the problem with all of it. The wins total roughly $1.5 billion. The losses totaled $6.9 billion. And the entire recovery now rests on a piece of technology. WPP Open is genuinely the most ambitious agentic AI architecture any holdco has built on paper. The Google Cloud deal is real. The InfoSum federation is real. The staff-adoption figure is real. And none of it is winning the pitches the company is losing.

One source described the WPP problem in a single sentence so clean it reframes the entire "turnaround" narrative. We are holding it for Part Two, because it names the actual failure point, and it is not the one the official story admits to. What we can say in the open is this: the "Build 2027, Accelerate 2028" timeline is not a recovery plan. It is a request that clients and investors hold through two more years of decline on the promise that the architecture eventually converts. Architecture is not a pitch win. WPP's own retention number says so. We will put that retention number next to Omnicom's in Part Two. They are pitched as peers. The numbers are not in the same universe.

Omnicom is the more interesting story, because Omnicom is the company that sounds like it is winning and is not.

Give them this. Omnicom has the best communications operation of any holding company, full stop. The IPG merger made them the largest holdco by revenue, roughly $25 billion pro forma. They relaunched Omni at CES 2026 with a polished narrative about ArtBotAI, Omni Assist, and "Connected Capabilities." They doubled the synergy target to $1.5 billion and announced a $5 billion buyback, and the stock jumped more than 15 percent. They own Acxiom Real ID, which on paper is the strongest pure US deterministic identity asset in the industry. The decks are immaculate. The releases land. The narrative is tight as a drum.

Now look at what they actually carried from idea to signed client.

The wins Omnicom leaned on coming into 2026, IBM, GSK, John Deere, Acadia, Little Caesars, Baileys, are mostly retentions and expansions, not competitive takeaways from Publicis. The biggest pitch of the entire cycle was Mars. Omnicom kept a slice of creative. Publicis took the $1.7 billion media prize off the table. All four holdcos touched one account, and the front-runner got the part that mattered.

That is the Omnicom problem in miniature, and it gets worse under the hood. The Acxiom Real ID asset is largely third-party referential data, and buyers privately rate it below Epsilon's transaction-based first-party graph. The integration that was supposed to weld Acxiom into Omni and produce a single connected stack has not delivered what the narrative promised. The data plumbing that looks seamless in a deck has not, by multiple accounts, shown up as seamless in a pitch. We have a source line on exactly how that integration is going, from someone in a position to know, and it is not flattering, and it is in Part Two.

And then there is the single most revealing decision Omnicom made all year. They are refusing to publish full-group organic growth in 2026. They will disclose only a "core operations" figure of 3.9 percent that conveniently excludes the dispositions. There is a four-word reaction to that decision from inside the building that we are saving for the paid section, but you can probably guess the shape of it. Companies do not hide good numbers.

So Omnicom is not losing the way WPP is losing. Omnicom is losing the way a company with great PR and a stalled integration loses: quietly, while sounding confident, coming second to Publicis in the rooms that decide the next decade.

Dentsu did the most honest thing of the four. It stopped pretending.

A record ¥327.6 billion net loss. A ¥310.1 billion goodwill impairment. A suspended dividend. A new CEO, Takeshi Sano, in the chair on March 27. The international business is in retreat across every region: APAC ex-Japan down 6.8 percent, Americas down 3.0 percent, EMEA down 1.8 percent. They lost Mastercard, a ten-year Carat client. The international entity count has been halved.

The tell is the CEO choice. Dentsu promoted a Japan-business veteran, not an international hire. That is not a failure of imagination. It is a strategy, and a clear-eyed one. Dentsu is consolidating around its genuinely excellent Japan business and quietly conceding the global game. One source framed it without flinching, and that quote, too, is in Part Two, because it is the most sympathetic and the most brutal thing anyone said about any of the four. The short version: Dentsu is no longer trying to be a Big Six holdco. The market will read that as failure. It is actually the only realistic plan on the table.

Which brings us back to the front-runner, and to the question every rival is privately asking. Why does Publicis get all the headlines?

The easy answer is the LiveRamp deal, the $2.2 billion acquisition in May that gave Publicis a third independent identity asset on top of Epsilon and Lotame. But the LiveRamp deal is a symptom, not the cause. Publicis gets the news because Publicis keeps doing the thing that generates news, which is winning the accounts everyone else is pitching. Plus 5.6 percent organic. 18.2 percent operating margin, best in the industry. €2.0 billion free cash flow. Seventh consecutive year of outperformance. Mars, Coca-Cola NA, and Paramount all taken directly from WPP.

Sadoun's framing is the cleanest articulation of the thesis any holdco CEO has produced. "In the age of AI, the name of the game is connect or die." The uncomfortable truth the other three are living with is that Publicis said it first, built the stack to back it, and then won the pitches that proved it. One company is executing. One is integrating and hoping. One is diagnosing and stabilizing. One is triaging and retreating. That is not a four-horse race. It is a procession with three explanations trailing behind it.

Here is the mic drop, and it is not ours. It is a number. In 2025, of every three dollars of media that changed agencies in the entire world, one of them went to Publicis. The other three global holding companies, the ones with century-old names and combined headcounts north of 300,000 people, split the rest while one of them went backward. That is not winning a cycle. That is rewriting who the competitive set even includes.

So here is the question that should be sitting on every CMO's desk and every procurement screen before the next holdco review. Three of the four big holding companies are running multi-year restructurings at the same time, and only one is actually winning. What does your MSA say about the integration risk you are about to sign into?

Your holdco is going to tell you the restructuring will not touch your account. There is a number that tells you exactly how much to trust that promise, and we put it in Part Two next to the one number that should make every WPP client read their contract this week. One holdco told its clients nothing would change and then kept a sliver of its own book. We will show you the sliver.

Every piece of coverage on the LiveRamp deal framed it the same way. Publicis bought identity infrastructure to feed its AI. Connect or die. The third leg of the Epsilon-Lotame-LiveRamp stool. All true, and all of it treats the acquisition as a thing Publicis did for itself.

Here is the reading nobody put in print. Ask not only what Publicis gained, but what it made sure no one else could have. Acquisitions at this altitude are rarely only about the asset. They are also about the board you are clearing. And on that board, the player who comes out worst is not Omnicom, which reacted loudly and has Acxiom to fall back on. It is WPP, which said almost nothing, and which had the most to lose precisely because of the bet it made twelve months earlier.

We want to be careful here, because we cannot read Arthur Sadoun's mind and we are not going to pretend to. Publicis has not said the deal was aimed at anyone, and there is no filing that frames it as a block. What we can do is lay out the structural case, because the structural case is real whether or not the intent was, and a CMO sitting inside WPP should understand it either way.

Start with the architecture WPP committed to. In April 2025, WPP acquired InfoSum, a clean-room platform whose entire founding premise is the rejection of the LiveRamp model. InfoSum was built on patented "non-movement of data" technology, the federated idea that you collaborate without onboarding anything to a central graph. Its leadership has gone out of its way to attack LiveRamp by name, calling the centralized data-broker model "very 2000, not very 2025," and stating flatly that "LiveRamp is not Switzerland." WPP did not just pick a different tool. It bet the whole house on the proposition that the centralized identity graph was the past and federation was the future.

Then Publicis spent $2.2 billion buying the centralized identity graph, at a 29.8 percent premium, in cash, and the market treated it as the defining move of the cycle. Whatever the intent, the effect on WPP is brutal and specific. It is not that Publicis took something away from WPP. It is that Publicis validated the exact architecture WPP bet against, at the exact moment WPP could least afford to have made the wrong call. The most-used identity layer in the open web is now owned by a competitor and growing, while WPP's federation bet looks, fairly or not, like the side of history that lost.

The case for the blocking-move reading is straightforward. LiveRamp's RampID connects more than 21,000 publisher domains and 500-plus partners. It is the closest thing the open web has to a universal identity currency. Any holdco that owned it would own a structural advantage over the other three, and removing it from the board denies that advantage to everyone else permanently. If you are Publicis and you can afford it, you do not just buy the best asset. You buy the asset whose absence hurts your rivals most. That is not conspiracy. That is how strategic M&A works at the top of a consolidating market.

The case against is equally honest, and we are not going to bury it. WPP was never a likely buyer. It had just spent on InfoSum and built its entire differentiation on the opposite architecture, so owning LiveRamp would have meant running two contradictory platforms. The rival with the actual live LiveRamp dependency was Omnicom, through the Acxiom relationship, and Omnicom is the one that reacted publicly. If this was a block aimed at anyone, the cleaner target was Omnicom, not WPP. So the "blocking WPP" reading is probably too neat. The likelier truth is that Publicis bought LiveRamp for its own thesis and WPP simply happened to be standing in the worst possible spot when it landed.

But here is why the distinction barely matters to a WPP client. Intent does not change exposure. Whether Publicis aimed at WPP or simply did not care about WPP, the outcome on a WPP advertiser's desk is identical: the identity infrastructure your agency bet against is now owned and accelerated by your agency's strongest competitor, your agency's owned alternative is smaller and entangled with that same competitor through shared clients, and your agency's recovery plan asks you to wait until 2028 to find out whether the architecture bet pays off. You are not buying a theory of Sadoun's motives. You are buying a contract into the consequences. That is the part the trade press missed while it was busy quoting the press release.

Part One told you Publicis won. Part Two shows you the receipts, and the receipts are the part nobody on a holdco payroll wanted on the record.

The quotes. We spent a month getting senior people across all four holding companies to talk with their names off, and the moment the names came off, the corporate script died. The data executive who explained, in one sentence, exactly why Omnicom keeps losing the rooms it should win. The WPP insider who described his own company's problem so precisely it should be the headline of their next earnings call, and never will be. The four-word reaction to Omnicom's decision to hide its organic-growth number. The most sympathetic and most brutal thing anyone said about Dentsu all year. None of it is in Part One. All of it is in Part Two. These are the things people say when they know you will protect them, and you cannot get them anywhere else, because everywhere else makes them use their names.

The scorecard. The full nine-dimension counterparty risk grid, all four holdcos rated side by side, the same framework that scored the Trade Desk and the verification duopoly. You have seen us hint that WPP scores worst and Publicis best. You have not seen the grid. The grid is the thing you screenshot and drop into a board deck. The grid is the thing that ends the meeting. WPP and Dentsu score a six on a scale where a clean vendor scores one or two, and the color-coded version of that fact is the single most useful page we have ever published for a procurement reader.

The bet WPP got wrong. The story the entire trade press missed while it was quoting the LiveRamp press release. Why WPP bought federation right before the company across the street spent $2.2 billion proving the centralized graph still wins, and the half-century of data heritage that explains why WPP cannot simply buy its way out. We are the only publication that has connected the 1969 fact to the 2026 fact. It reframes the whole deal.

The contract language. This is the part that pays for the subscription ten times over. The holdco-by-holdco decision tree, the exact clauses to demand depending on whose paper you are about to sign, and the one sentence to add to any holdco MSA in 2026 that closes the audit-and-restructuring loophole before they reach for it. If you are renewing with any of these four this year, the cost of not having Part Two is measured in the clauses you did not know to demand.

Here is the uncomfortable math. Three of the four big holding companies are running multi-year restructurings right now, and only one is actually winning. Your holdco has already told you, or is about to tell you, that none of this will affect your account. WPP told its clients exactly that and kept 16 percent of its book.