Pace Notes

On Tuesday ADOTAT reported that Nicolle Pangis, Netflix's vice president of advertising for the U.S. and Canada, is expected to leave, per multiple people familiar with the matter. Neither Netflix nor Pangis has confirmed it. Three days later, ADOTAT's follow up asks the better question: what would her exit tell you? Count the chairs. Jeremi Gorman, who built the ad business, left in October 2023. Peter Naylor left in 2024. Julie DeTraglia, measurement, left in 2025. Jon Whitticom, ads product, left in August. Pangis would make five senior ad leaders out since the ad tier launched in November 2022. Every exit got a gracious statement. The org chart never got one.

The org chart is the argument. A longtime ad tech executive told ADOTAT Pangis walked into "a separated ad product org that does not report to sales." In that setup, sales owns the number and product owns the timeline. In October Netflix cut its 2026 ad forecast from $3 billion to about $2.7 billion and blamed delays in product rollouts and partner integrations. Not sales. Weeks later, the sales chief is the one expected out. And on Nov. 2 Netflix hired Darshan Kantak from Google to run its ads platform, reporting to Amy Reinhard alongside the sales chief, not to her. New name, same box.

Then the pipes. Pangis was the public face of making Netflix easy to buy: The Trade Desk, Google's DV360, Yahoo and, since September 2025, Amazon's DSP. The price history ADOTAT lays out: about $65 CPM at launch in 2022, roughly $39 to $45 in 2023, and by this spring reports put it in the low $20s, with about half of non-live ad revenue running through outside DSPs. Every pipe made Netflix easier to buy. Every pipe also made it cheaper. One person told ADOTAT that The Trade Desk, fresh off its first revenue decline as a public company, is trying to recruit Pangis to run CTV sales and push Netflix toward an exclusive. That is one person's account, and ADOTAT labels it that way.

For buyers, a personnel change is a price change. Every upfront on Netflix's book was negotiated with a team Pangis ran, in what Digiday has called a buyer's market. ADOTAT's advice: get current terms in writing, ask who owns the account on Jan. 1 and who that person reports to, and price the same audience direct and programmatic side by side. The tell is the replacement. If the next sales leader still sits beside product under Reinhard, expect a sixth departure. The chair is not the problem. It is just the only thing Netflix keeps replacing.

Tal Jacobson, Perion's CEO, told AdExchanger that Affle lets every company it buys "stay independent under its own name." True, says ADOTAT, and that is the problem. In January 2021, Affle's Dubai free-zone subsidiary, Affle MEA FZ-LLC, bought the business assets of an Israeli company called Discover Tech, brand name included: $1.15 million up front, plus a success fee of up to $3.37 million. In October, Citizen Lab tied a company called Discover-Tech to ads that delivered Candiru spyware to at least 45 devices at Al Jazeera. Citizen Lab also says it has no reason to believe Affle controls those operations. Perion keeps its name. So did the last guy.

The math from Monday still frames everything. Affle pays about $420 million for a company holding $228.3 million in cash, so roughly $192 million for the operating business, one Perion spent about $685 million building. The family photo explains the discount: SweetIM add-ons that Malwarebytes still flags, Conduit's toolbars, Undertone, which by its own account invented the pop-under, and about 80 clickbait sites shut in under two days once Sincera looked. On Affle's side: SMS 2.0, which showed a full-screen ad as your text sent; Appnext, which sells apps preinstalled on new phones; and a November 2024 SEBI warning over undisclosed related-party transactions with its own subsidiaries. After closing, Perion is one of those subsidiaries.

Then the useful part: twelve questions for Perion, Undertone and Hivestack reps, with the dodge you will hear and the reply that makes it cost something. Perion paid $55.9 million in traffic acquisition costs last quarter, 57% of revenue, so ask for the top ten recipients on your campaigns, by name, and check the seller IDs yourself. Ask whether your CPM deal turns into pay-per-conversion, because whoever counts the conversion gets paid for it. Ask for every domain and app with no "other" line. For Hivestack screens, ask for proof-of-play logs, dark screens included, and who audits the audience multiplier.

ADOTAT also hands buyers three clauses: full supply-path disclosure within ten business days, a pricing-model lock with conversions counted by a partner you pick, and the right to walk without penalty within 90 days of a change of control. The window matters. The go-shop ends Dec. 11 and closing is expected early next year, after which Affle has told investors it will review Perion's contracts and its revenue recognition. Ask now, while the person across the table still needs your renewal.

On Monday, OpenAI's head of monetization, Benji Shomair, said trust guides its ads business, and that partners sometimes ask for things OpenAI "just can't" do. The same week, ChatGPT's new "visual ads" started appearing in the pause while it generates an image you asked for, and OpenAI opened a self-serve Business Hub in beta so brands can see how people engage with them inside ChatGPT. The trust talk and the new ad slot arrived together.

Now the receipts. A working paper by Maximilian Kaiser and Christian Schulze, still under review, looked at U.S. traffic and found a paid ChatGPT visit earned $0.89 per session, against $2.45 for a paid Google Search visit. That is about 64% less. Paid ChatGPT traffic converted at 0.37%, against 1.17% for Google. Among 87 U.S. sites buying both, 97% got less revenue per visit from ChatGPT. By the authors' math, a ChatGPT click is worth about 26 cents against a typical 72-cent search click.

Why it matters: buyers are being asked to fund a new channel on the strength of a story about the user. The paper suggests the user arrives less ready to buy. That does not make ChatGPT ads worthless, and one working paper is not the market. But it gives every media planner a number to put next to the pitch, and it lands while OpenAI is still adding formats. ADOTAT has already flagged the other half of the bargain: when AI tools rewrite your ad for you, the advertiser still carries the liability for copy nobody on its side read.

ADOTAT's line on the whole week fits here too: companies told you what they were, and their own paperwork disagreed. OpenAI calls the product trust. The early receipts call it about a third of a Google click. Before shifting search money, ask OpenAI for conversion data on your own campaigns, measured by someone you pick, and price the click accordingly. Trust is the pitch. The invoice is the test.

The Coalition for Innovative Media Measurement published a study on Nov. 12 called "Behind the Bidstream," and PPC Land's Luís Rijo walked through it the next day. The question is simple: what happens when a CTV ad request shows up missing the signals a buyer needs to judge it? The answer, from 50 programmatic CTV buyers: 62% shift spend to another supply path, 54% broaden targeting and 46% lower their bids. Only 8% said their DSP always acts on the signals they require. The pipe carries the data. Nobody promises the pipe reads it.

The study's sharpest number is about verification, not targeting. Asked how likely they were to bid, 88% scored high on detailed signals over a verified path, 66% on limited signals over a verified path, and only 36% on detailed signals over an unverified path. Supply-path transparency made the top three factors for 48% of buyers, edging out scale and reach at 46%. Buyers will take less data from a known seller over more data from a stranger. The study was co-sponsored by FreeWheel and OpenX, which also supplied bid-log analyses.

CIMM's fix is a voluntary minimum trading baseline, built on inventory identity and selling authority, plus platform capability statements, reconciliation reports and an industry pilot. There is no timetable, no named participants and no start date. And the limits are real: 50 respondents is a small sample, the bid logs show associations rather than causes, and, as PPC Land notes, the DSP side is not among the sponsors. CIMM's managing director Jon Watts put the goal plainly: "Publishers need to understand what buyers require to assess their inventory."

This is the same plumbing ADOTAT keeps opening up. Netflix built four DSP pipes and watched its CPMs fall as buyers learned to shop them. ADOTAT's Field Kit tells Perion's buyers to demand every seller ID and check it against sellers.json themselves. And its LiveIntent reporting found a bot getting an identity from 17 of 18 ID vendors. CIMM's buyers are saying the same thing in survey form: when the path is unknown, the money leaves. A baseline nobody has to sign is a suggestion. Buyers are already enforcing their own.