Pace Notes

When Fox agreed to pay $22 billion for Roku on June 15, Lachlan Murdoch told analysts what he was getting: a home screen he called beachfront property, and what he called Roku's unique expertise in performance marketing. The deal closed Friday, Nov. 6, months early. That same morning a staff memo said Charlie Collier, who ran Roku's ad and platform business for four years, was not coming along. Neither was general counsel Chris Handman. Roku's ads now report up to Mustafa Ozgen, the executive who ran its devices and technology. Fox kept the expertise it praised in June off the chart.

Roku's own filings explain why the sales floor is nervous. Third-quarter ad revenue rose 23% to $727.4 million. But Roku served 33% more video impressions and earned 7% less for each one. In the first quarter it was 59% more for 14% less. Across nine months, impressions rose 42% while the average price fell 11%. That is growth by volume, not by price. Fox's closing release put about $400 million a year in cost savings on a two-year clock, and people at Fox told ADOTAT Roku's PR and marketing team will likely be cut. One person who knows the ad business said Roku's sellers "took whatever" they could get.

The pitch Fox inherited has a proof point that does not prove what it says. Roku's sales script cites a 40% drop in cost per action from an Omnicom study. ADOTAT read the study. It was a survey test with roughly 300 people per group, measuring answers like intent to search. No purchases. No live campaigns. No cost per action. And the grader was a holding company that buys the inventory.

Then there is the drug pitch. Roku has been selling health marketers targeting built on what the TV recognizes you watched, home screen units, QR codes, retargeting and marketing to doctors. ADOTAT sent Roku five questions on Oct. 28, starting with whether viewers consented to that use. Roku did not answer. Neither did DeepIntent. Michigan's attorney general is already suing Roku over data, and an FDA rule on drug ads has a December target. If you buy Roku, get your team, your data access, your measurement and your price history in writing this month. Fox paid for an expert. Ask it for the name.

Perion describes itself as an advanced technology leader. Its chairman, Eyal Kaplan, says the Affle sale gives shareholders compelling value. Here is the value. Affle pays about $420 million. Perion had $228.3 million in cash and no debt on Sept. 30. Back that out and the business sells for about $192 million, roughly 3.7 times EBITDA guidance Perion has since withdrawn. Verve's 2024 deal for Jun Group was called a bargain at 4.7 times. Perion spent roughly $685 million on acquisitions and its own stock. The market priced all of it below the cash it came with.

So where did the cash come from? For most of Perion's life, from one customer. The filings call it Customer A, and the numbers point to Microsoft: 81% of revenue in 2015, 51% in 2020. Perion routed searches to Bing and took a cut. When Microsoft cut publishers out of its marketplace in 2024, Bing went from 35% of revenue to under 5%. Last quarter Perion paid 57% of revenue straight back out in traffic costs. First-half adjusted EBITDA was $3.2 million. The bank balance is what is left of the Bing years.

The family tree is in the filings too. SweetIM in 2012, browser add-ons that Malwarebytes still flags as "PUP.Optional.Perion." Conduit's toolbars in 2014. Undertone in 2015 for $180 million, a company that says it invented the pop-under. Content IQ and Pub Ocean in 2020, about 80 clickbait sites that Perion shut in under two days once Sincera looked. Pesach, who ran a competing ad network then, remembers Undertone running with the adware crowd, Zango included, which the FTC charged in 2006 and which paid $3 million to settle. Every branch lives on someone else's traffic.

The deal itself has a timeline worth a look. The merger agreement cites an Affle and Perion confidentiality agreement dated Feb. 9, nearly nine months before the announcement. In March, Canaccord started coverage of Perion at Buy with a $14 target. By November it was Perion's adviser, selling at $11.70. The filings do not say when Canaccord was hired. The go-shop ends Dec. 11. If your money runs through Perion, get your rates in writing, ask whether your deals become pay-per-conversion under Affle, and ask for a traffic-source report. If nobody can produce one, that is your answer.

By Nielsen's count, college football games are averaging 1.6 million viewers through Week 9, down 11%, across 402 games against 410 last year. Michael Mulvihill, who runs analytics at Fox Sports, says total minutes viewed are up 2%, because more games are on national TV and the volume, in his words, negates the lower average. Negate means cancel. It does not mean add.

Do the multiplication. An average down 11% across eight fewer games should put total viewing down roughly 12.7%, by Unmeasured's arithmetic. For both numbers to be true on the same games, broadcasts would have had to get about 17% longer in one season. Nobody has said that. The likely answer is that Fox is counting a different slice: the big five networks aired 131 games through Week 8, up from 106 windows in 2021. And per The Athletic's Stewart Mandel, Fox is the only one of the five that is up. The number comes from the one network it flatters.

The decline also survived a more generous ruler. Nielsen has changed its method three times in under two years, most recently a co-viewing change on Aug. 31. Mandel puts the latest lift at about 3.4%. On last year's ruler, the sport would be down closer to 14%. The top is fine: top-10 matchups are up 6.4%. The middle is not. Games between teams ranked 11 to 15 and those below are down 22.7%. That middle tier is exactly what the 24-team playoff is supposed to sell.

The NFL, meanwhile, called Nielsen's changes "overly rushed" in August. Its viewership is now up 3%, within what the method change alone could explain, and the league credited close games. For a buyer the rule is simple. The average is your number. The total is the rights holder's number. Buy the game, not the season, ask what the new ruler added to your delivery, and do not let total viewing set your CPM.

On July 16, co-CEO Greg Peters told investors the FIFA game and Unhinged were Netflix's best cloud game debuts, "top tier." On July 31, the studio behind the FIFA game, Refactor, laid off about 85% of its staff, a cut its part-owner Delphi made, not Netflix. On Aug. 13, 28 days after the call, Netflix closed Night School, which made Unhinged, and Moonloot, which never shipped a game. That is how Netflix leaves a business. Quietly, and on schedule.

Then came the numbers nobody was meant to see. Netflix cut about 1,000 jobs on Oct. 14, and a LinkedIn list of displaced games staff ran to nearly 60 names. The Wall Street Journal reported an internal memo: only 3.5% of subscribers play in a given month, short of Netflix's own target, and the company had stopped pushing games. The next shareholder letter skipped games entirely, apart from a Grand Theft Auto VI sneak peek.

The reason is the scoreboard. Netflix grades itself on viewing hours, and a game hour is not a watched hour. Look at what survived: party, kids' and TV games like Boggle and Pictionary, played on the couch with the TV already on. What died was everything that pulls you out of the show. And games never carried a single ad. Ads in games were debated in 2024 and never shipped. Now Netflix has cut its 2026 ad forecast to $2.7 billion from $3 billion, and its share of U.S. TV viewing slipped to 7.8% while YouTube's climbed to 14.2%.

A product that eats viewing time and produces zero ad inventory is not a growth bet. It is a leak. The one deal that could have made Netflix a real publisher was Warner Bros., games studios included. Netflix walked away in February with a $2.8 billion breakup fee, and Paramount got the studios. Netflix wanted to be a games publisher. It settled for a games shelf.

The Trade Desk reported third-quarter revenue of $736 million on Monday, against $739 million a year earlier. That is a 0.5% decline, and PPC Land's Luís Rijo notes it is the company's first since its 2016 IPO. A year ago the same quarter grew 18%. Net income fell to $38 million from $116 million. Adjusted EBITDA came in at $266 million, down from $317 million.

Jeff Green called the quarter "well ahead of our expectations entering the quarter." True, and that is the trick. In August the company guided to at least $650 million. PPC Land's arithmetic puts the beat at about $86 million. Clearing a bar you lowered is not the same as growing. The release also books $44.3 million in restructuring charges, mostly severance, kept out of adjusted EBITDA. Fourth-quarter guidance is at least $755 million, which PPC Land puts 11% below the same quarter of 2025. The board added $731 million to the buyback, for $1 billion available.

The line to circle is about Netflix. The release says an expanded integration starts this month, bringing Netflix's first-party and third-party data into Kokai, with measurement to follow. ADOTAT reported Tuesday, on one person's account, that The Trade Desk is trying to recruit Nicolle Pangis, Netflix's U.S. and Canada ad sales chief, to run its CTV sales, and wants to use her to push Netflix toward exclusivity. ADOTAT has not confirmed talks. But Netflix also sells through Amazon's DSP, the deal that sent Trade Desk shares sliding in 2025.

So the quarter tells you where The Trade Desk thinks growth lives now: in someone else's premium inventory. ADOTAT's Nov. 6 Wire flagged that guidance implied the first decline as a public company. The decline arrived, smaller than feared. The question for buyers is whether "open internet" still describes a company courting one streamer for an exclusive.