IndustryPace
Ad industry commentary from Pesach Lattin. Since 2010.

Pace Notes

Thursday, November 19, 2026

The ad business, read one turn ahead.

Four ADOTAT stories and one from outside, and every one is about who gets to check the count. HUMAN caught a fraud ring nobody else claims to have seen, Samsung hired a Google seller to sell what its TVs watch, Remark ranks its own numbers honestly, and Criteo is taking calls. The Pick is BADBOX 3.0.

From ADOTATHUMAN Says Its Customers Were Protected. BADBOX 3.0 Was 4% of the CTV Traffic It Saw.

HUMAN Says Its Customers Were Protected. BADBOX 3.0 Ran Through Their Pipes.Pick

By ,

On Nov. 18, HUMAN Security announced it had disrupted BADBOX 3.0, which at peak pushed 40 billion bid requests a day into programmatic. That is more than three times VASTFLUX, the old record. But the number ADOTAT pulled out of the technical report is quieter and worse: at its height the operation was more than 4% of all the connected TV traffic HUMAN observed. HUMAN does not run an exchange. It scores requests its paying SSPs, DSPs and platforms send it. One in 25 CTV requests HUMAN saw was already inside the pipes.

The mechanics are cheap and old. Off-brand, uncertified Android boxes built in mainland China ship with a backdoor. The box pulls ad modules from public code repositories and plays real video ads in a window nobody can see, while the bid request names a well-known CTV app that never ran. One method in the code is called showHideWindowAd. The same boxes were enrolled in eleven residential proxy networks, and HUMAN ties the module catalog to a Chinese maker, TXCZ. Most damning for the buy side: the malware sent hardcoded values so an ad server could credit the operators. Somebody's ad server knew whom to pay.

The tells were not hidden. HUMAN lists them: a handful of box models aimed almost entirely at one family of apps, about 80% of one app's U.S. requests arriving from residential proxy exits, different apps sharing IP addresses, a spike at the same hour every day, a test-build browser string on a television, and senders missing from the apps' sellers.json and app-ads.txt. Every one of those sits in a log an exchange already holds. ADOTAT's line is the right one: that is not reverse engineering, that is a database query. And the family has a calendar: BADBOX in 2023, BADBOX 2.0 at 750 million a day in 2025, a public warning from HUMAN that a third was coming, and now this.

Why it matters: ADOTAT checked the public record for DoubleVerify, IAS, Pixalate, The Trade Desk, Magnite, Index Exchange, PubMatic, OpenX, FreeWheel, Amazon and Google. None has said it found BADBOX 3.0 itself. Some may have filtered it quietly, and several filter through HUMAN, which is the point. The MRC told ADOTAT that pre-bid vendors cannot block a bid; the platforms decide. So the industry has outsourced its fraud research to one threat team and bills on the result. ADOTAT's seven questions for an SSP or DSP start with how many requests reached you at peak and end with what are we paying you for that HUMAN is not already doing.

The full story is at ADOTAT

Expanded from ADOTAT's Nov. 18 investigation by Pesach Lattin; part two of the original sits behind an email signup. Per ADOTAT, its sources were HUMAN Security's Nov. 18, 2026 BADBOX 3.0 technical report and press release, HUMAN's earlier reports on VASTFLUX, BADBOX, BADBOX 2.0 and NewsJunkie, HUMAN's Magnite case study and partner announcements, Google's statement in the release, Nokia Deepfield's research, reporting by The Register and PPC Land, DoubleVerify and Pixalate research, Adalytics, and the Media Rating Council's written answers to ADOTAT in October 2026. HUMAN's figures have not been independently verified, a bid request is not a billed impression, and HUMAN published no device count, dollar figure, app names or seller names. ADOTAT searched the public record on Nov. 18, with Perplexity and ChatGPT search as cross-checks, for any platform claiming detection and found none; the absence of a statement does not prove a platform missed the traffic. Named companies had not yet been asked for comment; ADOTAT says responses will run as dated updates. Pesach Lattin ran an ad network in the late 2000s and has known executives at several companies named here for years. ADOTAT is reader-supported with no paid placement in editorial. The Trade Desk, Magnite, PubMatic, DoubleVerify, IAS, Amazon and Alphabet are public companies; this is commentary, not investment advice.

From ADOTATSamsung's TVs "Do Not Spy." Nelson Boyce Now Sells What They See.

Samsung Says Its TVs "Do Not Spy." Texas Made It Ask Before They Watch.

By ,

Nelson Boyce, eleven years at Google and most recently a leader of its U.S. travel ad business, is now senior vice president of ad sales at Samsung Ads, Adweek reported on Nov. 10. He told Adweek the company is at "a really opportune moment." The year says otherwise. Texas sued Samsung over automatic content recognition in December 2025. Samsung settled on Feb. 26 and agreed to get express consent from Texans before collecting ACR data. Samsung's line after that: its TVs "do not spy on customers." Boyce's job is to sell what they see.

The chair was empty in public for a long time. LoopMe announced on Feb. 24 that it had hired Michael Scott, Samsung Ads' North America sales and operations chief. ADOTAT counts 259 days from that release to Adweek's, and found no public name in the role between. Samsung's March NewFronts release quoted nobody in ad sales leadership. Meanwhile the U.S. parent moved its headquarters to Plano, Texas, the state that sued it, and a New Jersey notice listed 739 employees at the old campus laid off or offered relocation. Samsung upgraded the title from vice president to senior vice president. Harder jobs get bigger titles.

The product changed too. Samsung's March pitch was Performance TV, and its headline partner was Amazon Ads: add-to-cart ads on Samsung TV Plus, inventory bought through Amazon's DSP and measured in Amazon Marketing Cloud. So the performance story closes in a rival's cart, on a rival's pipes. What Samsung still owns is the glass, and the ACR data it calls "exclusive." That glass is now under a Texas consent screen, a proposed class action in New York and Tillery v. Samsung in California, where plaintiffs want the same opt-in. An ACR footprint sold on scale is a different product once the viewer has to say yes.

Why it matters: connected TV keeps hiring from Google because the budgets it wants are search budgets, and Boyce has sold to Priceline-type accounts for years. In 2024 he said consumers want a relationship "that respects their data." In 2023 he said "good input into AI is good output from AI." Samsung's input got thinner this year. ADOTAT's five questions for his first agency meetings start with the one Samsung has not answered: what is the opt-in rate in Texas? That number is now the rate card.

The full story is at ADOTAT

Expanded from ADOTAT's Nov. 18 report by Pesach Lattin. Per ADOTAT, its sources were Adweek's Nov. 10 report and exchange4media's account of the hire, Samsung and LoopMe press releases, a 2024 Boston Hospitality Review interview with Boyce, recordings of his 2021 Phocuswright and 2023 Skift Global Forum appearances, a 2023 podcast and a 2017 magazine profile, published accounts of the Texas lawsuit and settlement and of two proposed federal class actions, coverage of Alphabet's third-quarter call, and reporting by The Dallas Morning News, Reuters, The Information and AdExchanger. ADOTAT did not review the court filings directly; the descriptions of the complaints come from published summaries and are allegations, not findings, and Samsung admitted no wrongdoing in the Texas settlement. The 259-day figure runs from LoopMe's announcement to Adweek's report and does not measure how long the role was vacant. Samsung has not said whether Samsung Ads staff were among the 739, or what share of Texans opted in. ADOTAT is reader-supported with no paid placement in editorial. Samsung, Amazon and Alphabet are public companies; this is commentary, not investment advice.

From ADOTATNielsen Says Streaming Hit 49%. It Also Says Don't Sell Ads.

Simulmedia's Dave Morgan Says the Upfront Is Fading. Linear Still Took More of It.

By ,

On Nov. 18, Dave Morgan of Simulmedia argued in The Current that the TV upfront is giving way to an always-on forward market. Pesach Lattin, who calls Morgan a friend and one of the few people whose read he would stop to hear, answered on Nov. 19. He agrees on the destination. He disagrees on the clock, and on the instruments used to read it. His frame is worth copying: professionals disagree, show their work and let the other side answer. No insinuation, no ad hominem.

The first instrument is Nielsen's Gauge. Morgan cited streaming at 49% of TV in July. Accurate, and the most flattering month on the board, lifted by the World Cup and the usual summer bump. More important, Nielsen prints in every Gauge release that it does not reflect the currency ratings that inform ad sales. Morgan's figure of "nearly 52%" of ad-supported viewing for linear is not in the Gauge at all; the published shares sum to roughly 42% of total TV. ADOTAT thinks the derived figure is probably right in direction, and says it will print a correction if a Nielsen product backs it. The most-cited number in TV advertising carries a label saying not to use it for TV advertising.

The second instrument is money. The headline said streaming passed linear in the upfront. Media Dynamics' estimate is $17.2 billion for streaming against $16.6 billion for linear, in primetime, with roughly $12 billion more in linear outside primetime. In total, linear still took more. eMarketer had the same milestone at different numbers. And the market grew on volume, not price: primetime upfront near $33.8 billion, up about 9%, while streaming's adult CPM fell 4.9%. A continuous market sells the scarcity the upfront manufactures, and scarcity is what holds price up.

Why it matters: an always-on market needs one currency both sides compute against, live, with a tiebreak. The industry has four certified national currencies, an incumbent with an estimated 80 to 90 percent of volume that spent the spring answering the MRC's priority list, and no tiebreak. ADOTAT's bet is that automation arrives before agreement, because automation has vendors and agreement has committees. It also notes, without insinuating anything, that The Current is The Trade Desk's publication and the op-ed predicts a market run by integrated DSP and SSP platforms. The drinks, Lattin writes, are on him.

The full story is at ADOTAT

Expanded from ADOTAT's Nov. 19 opinion piece by Pesach Lattin, a response to Dave Morgan's Nov. 18 op-ed in The Current; as commentary, no comment was sought from the parties named. Per ADOTAT, Morgan's figures were checked against Nielsen's Gauge releases and methodology notes, the Media Rating Council's March 3 and May 27, 2026 status updates on Nielsen's Big Data + Panel service, the U.S. Joint Industry Committee's September 2026 recertification of Comscore, iSpot and VideoAmp, upfront estimates from Media Dynamics Inc and eMarketer as reported by MediaPost, Señal News and others, and the January 2026 NBCUniversal, RPA, FreeWheel and Newton Research announcement. Limits: the upfront dollar figures are estimates, not audited spend, and the two forecasters disagree; ADOTAT worked from published Gauge releases, not data files; the 80 to 90 percent share for Nielsen is a market estimate, not a disclosed figure; and the source of the 52 percent figure could not be identified and may come from a Nielsen product ADOTAT has not seen. The author and Dave Morgan are friends, and the author has known many of the executives named for two decades. ADOTAT says it has no financial relationship with Simulmedia, The Trade Desk, Nielsen, NBCUniversal, Media Dynamics or any JIC-certified provider, is reader-supported and takes no paid placement in editorial.

From ADOTATRemark Runs a 50-50 Holdout on Every Brand. Amazon Uses a Seven-Day Window.

Remark Pitched Chat-Attributed Revenue. Its CEO Says the Split Test Is What Counts.

By ,

A 5WPR pitch this fall said Remark, a Boston AI shopping advisor company, had four retailers generating or approaching $10 million in chat-attributed revenue, and in the same paragraph cited 8 to 12% revenue lifts. Two kinds of number, blended the way the industry blends them. CEO Theo Satloff pulled them apart himself, in writing. The lift comes from a 50-50 A/B test Remark runs on every new partner, which he called the "only way to truly measure incremental impact." The revenue figure counts any order within 30 days of a chat. The CEO ranked his own headline number second.

Set that next to the biggest shopping assistant on earth. Amazon said Rufus was on track for more than $10 billion in incremental annualized sales, then nearly $12 billion, reportedly on a seven-day attribution window, with no published holdout, control or method. In 2026 Amazon merged Rufus into Alexa for Shopping, stopped repeating the incremental billions, and started selling sponsored prompts inside the answer. From $12 billion in claimed lift to ads in the reply, with no homework shown on either.

The rest of the category mostly reports a window and a total. Gorgias earns credit for calling its own three-day number associative, not causal, and for disclosing that about 80% of influenced revenue rests on a chat simply happening before an order. Other tools let merchants choose 7, 14, 30 or 60 days, which ADOTAT likens to letting the defendant pick the jury. Remark's lift claims have sat at 8, 9, 10, 8 to 12 across two years and several surfaces. Invented numbers inflate. Control-group numbers get boring.

Why it matters: this is the year AI shopping surfaces started selling placement: Amazon's sponsored prompts, Google's Direct Offers in AI Mode, labeled ads under ChatGPT answers. Satloff's answer on ads is the cleanest ADOTAT has heard: "A recommendation is only valuable if it can't be bought." He also names the data fight: inside ChatGPT, brands only see the referral; on their own site, they keep the transcript. ADOTAT's checklist for any AI shopping vendor starts with show me the holdout and put the window in writing.

The full story is at ADOTAT

Expanded from ADOTAT's Nov. 18 report by Pesach Lattin, built on written answers from Remark CEO Theo Satloff provided through 5WPR in November 2026, plus Remark's case studies, funding announcements and launch materials, Amazon earnings statements, Gorgias's published methodology and trade coverage of OpenAI's and Google's 2026 ad and commerce products. Remark's revenue, lift and satisfaction figures are company-reported and were not independently audited; ADOTAT did not see the underlying A/B test data. Amazon's incremental figures and the seven-day window come from earnings statements and trade reporting, not published methodology. Remark did not review the story before publication. ADOTAT says it has no financial relationship with Remark, 5WPR, Amazon, Google, OpenAI or Gorgias. The author ran ad networks in the 2000s and was director of business development at LeadsCon, has known 5WPR for years, and ADOTAT has previously reported critically on 5WPR's founder. ADOTAT is reader-supported with no paid placement in editorial. Amazon and Alphabet are public companies; this is commentary, not investment advice.

ElsewhereReuters, via SRN News: Exclusive-AppLovin among firms in talks to acquire Criteo, sources say

Criteo Bet on Retail Media. Now It Is a $1 Billion Takeover Target.

By ,

On Nov. 17, Reuters' Milana Vinn reported, citing two people familiar with the matter, that AppLovin is among the firms in talks to acquire Criteo. Price and terms were unclear. A deal could be announced before the end of the year, the sources said, but none is guaranteed and Criteo could stay independent. Criteo declined to comment. AppLovin did not respond.

The backstory is a slow fade. Reuters reported in July that Vista Equity Partners and the smaller Quinti Capital had approached Criteo about a joint buyout, after which Criteo hired a bank to look at strategic alternatives, including a sale. Founded in Paris in 2005, Criteo has seen its shares fall more than 50% over five years, to a market value of about $1 billion, per LSEG, despite its turn to retail media. AppLovin was worth about $136 billion on Monday. The buyer is more than a hundred times the size of the target.

The fit, as AdExchanger read it on Nov. 19: Criteo plugs into retailers and web publishers and holds a large graph of shopper data, while AppLovin operates almost entirely in-app and has said adding ecommerce advertisers is a priority. AdExchanger also recalls that Amazon was once rumored to have looked at Criteo before building its own off-site retargeting. Criteo spent a decade selling access to shoppers. Now the shoppers' graph is what is for sale.

What ADOTAT has already written: in its Nov. 4 identity report, built on a peer-reviewed study, Criteo was among the 17 of 18 ID vendors that gave a self-declared research crawler an identity. Any buyer is paying for a graph, so ask what the graph refuses to identify. And ADOTAT's Nov. 18 Remark and Samsung reports describe the same fight from two sides: whoever owns the shopper conversation or the checkout owns the customer, and a TV maker borrowed Amazon's cart because it had none. AppLovin, a company built inside apps, appears to want a cart on the open web.

Read the original at SRN News

Commentary on Milana Vinn's Nov. 17, 2026 Reuters exclusive, with additional reporting by Mathieu Rosemain, as published by SRN News; the Reuters original was not read separately. The talks rest on two unnamed sources; no price was reported, and Reuters says no deal is guaranteed. Criteo declined to comment; AppLovin, Vista Equity and Quinti did not respond to Reuters. Market values are LSEG figures as reported by Reuters. The read on fit, the Amazon recollection and AppLovin's ecommerce priority come from AdExchanger's Nov. 19 daily roundup. The ADOTAT references in the last paragraph come from the linked ADOTAT reports; ADOTAT has not reported on these talks. AppLovin and Criteo are public companies; this is commentary, not investment advice.