Ad platforms are judged on what they do when they work. The last few days of August offered a look at the other condition. A rewarded video format that pays only on completion stopped completing on Android for the better part of eight days, and the failure was filed under the mildest label Google's status dashboard offers. Two brand safety controls in Display and Video 360 were scheduled for removal with no migration window attached. Criteo lost the executive who owned every product it announced this year, five weeks into a private equity approach it has not resolved. And Amazon put a number on the part of its European business that exists because customers send things back.

The four stories share a subject, which is maintenance: the unglamorous work of keeping systems running, keeping controls available, keeping a product organisation staffed, and turning returned goods back into sellable stock. None of it is announced with a keynote. All of it determines what an advertiser or a publisher can actually do in a given week.

Eight days of a static play icon

Between August 20 and August 28, 2026, rewarded video advertisements failed to play inside Android applications served by Google's AdMob. Users tapping to watch saw a static play icon where the video should have started, and PPC Land reported on August 29 that the failure ran for roughly 188 hours from the first dashboard notice to the most recent update.

The timeline is worth reading in sequence, because the gaps in it carry more information than the entries.

Google logged the incident at 21:19 UTC on August 20, with the first dashboard notice appearing five minutes later. The second update came at 01:10 UTC on August 22, twenty-eight hours after the first, and it indicated partial restoration while stating plainly that no workaround was available. Then nothing. The third update arrived at 17:16 UTC on August 28, after roughly 160 hours of silence, and reported that engineering had identified the root cause while describing mitigation as covering most users rather than all of them.

Rewarded video is not a marginal format. The mechanism is unusually literal: a user agrees to watch an advertisement to completion in exchange for an in-application reward, a life, a hint, a currency top-up, and the publisher is paid on that completion. Break the playback and the transaction has no settlement step. An impression that never renders cannot be completed, and a completion that never happens cannot be billed to the advertiser or credited to the developer. The format is the closest thing mobile advertising has to a fully consented exchange, and it is the format that stopped.

The absence of a workaround was structural rather than an oversight. Rendering is controlled by the Google Mobile Ads SDK, compiled into the application binary and shipped through the Play Store. Nothing in an AdMob account settings panel governs whether a video element initialises. A publisher who noticed revenue falling on August 21 had no configuration to change, no mediation waterfall adjustment that would have helped, and no path to a fix that did not run through Google's own engineering queue.

What makes the episode legible as a governance question rather than a bug report is the classification. Google's dashboard marked the incident as service information, the lowest disruption tier available, a label ordinarily used for changes that do not interrupt revenue. For eight days the tier said one thing and the format did another.

The disclosure gaps are equally specific. Google's dashboard published no impression volumes, no revenue impact estimate, no geographic scope, no count of affected publishers, no statement on whether iOS applications were touched, and no answer to the question that matters most for reconciliation: whether impressions were billed to advertisers despite the completion event never firing. An advertiser buying rewarded inventory on a completed-view basis and a publisher reading a payout report during that window were looking at the same eight days from opposite sides, and neither has been told what the numbers mean.

This is not the first infrastructure failure of the year to land on the same set of publishers. In January 2026, Google Ad Manager and AdSense suffered simultaneous disruptions beginning at 18:00 UTC on January 13, with a second incident on January 14, and Google acknowledged systemic Ad Exchange match rate declines at 01:44 UTC on January 15. PPC Land documented earnings declines of 50 to 90 percent within twenty-four hours, including a 64 percent RPM decline on German sites, 82 percent on .com domains, 90 percent on Spanish-language sites and 70 percent for Swiss publishers. One multi-site operator wrote at the time that they had never seen figures like it.

The financial context sharpens the point. Alphabet's Google Network segment, which contains AdMob, AdSense and Ad Manager, has been contracting while the rest of the advertising business grows. PPC Land reported that Network revenue fell 4 percent to 6.97 billion dollars in the first quarter of 2026, and the second quarter brought a further 1 percent decline to 7.3 billion dollars. Search advertising over the same period rose 17 percent to 63.3 billion dollars. A segment shrinking while its sibling grows by double digits is a segment competing for engineering attention it does not obviously win.

Publishers on the platform are also mid-migration. Google has required Android developers to move off the legacy Mobile Ads SDK by 2028, a rewrite that changes initialisation, ad loading and format handling. An outage rooted in SDK-controlled rendering, occurring during a mandated SDK transition, raises a question about which code path failed that Google's three dashboard entries do not answer.

Mediation offers no escape either, and the reason is worth stating precisely. A publisher running AdMob mediation with several demand partners can shift fill toward a different network when one underperforms, and many do so automatically. But mediation arbitrates which network wins an impression; it does not change which SDK draws the creative once AdMob has won. If the winning bid comes back from AdMob and the render fails, the waterfall has already resolved. The auction worked. The video did not. Any developer watching only fill rate and eCPM would have seen a healthy auction sitting on top of a broken delivery layer, which is precisely the failure mode that a reporting dashboard built around fill and eCPM is least equipped to reveal.

There is a second-order consequence for anyone reconciling August. Rewarded video sits at the top of most mobile eCPM tables because completion is near-guaranteed by design, so a period in which it silently stops serving pulls a publisher's blended eCPM down through composition alone, without any single line item looking obviously wrong. Attribution of an August revenue dip to seasonality, to creative fatigue or to a mediation change would be a reasonable conclusion drawn from the available data and, for these eight days, the wrong one.

Two exclusions leave DV360, and the notice was unversioned

The AdMob failure was accidental. A notice published the day before was not.

On August 27, 2026, Trevor Mulchay of the Display and Video 360 API team posted a change notice on the Google Ads Developer Blog, and PPC Land reported that two brand safety exclusion controls disappear from DV360 on October 1, 2026.

The first is digital content labels. Advertisers will no longer be able to exclude specific labels through the API targeting type TARGETING_TYPE_DIGITAL_CONTENT_LABEL_EXCLUSION, nor through the corresponding Digital Content Labels - Exclude column in structured data files. Digital content labels are the maturity ratings that let a buyer keep a family brand out of inventory rated for older audiences, and they have been part of the standard exclusion vocabulary for years.

The second is sensitive categories. Google's notice states that the majority of sensitive categories can no longer be excluded using targeting, removing TARGETING_TYPE_SENSITIVE_CATEGORY_EXCLUSION and its structured data file equivalents. The notice does not specify which categories survive, which leaves buyers to discover the surviving list by testing rather than by reading.

Two mechanics in that announcement matter more than the categories themselves.

The changes are unversioned. They apply to every supported API version simultaneously, without the migration window that a versioned deprecation would provide. A buyer running on an older API version gets no reprieve, because there is no older behaviour left to run on. Anyone maintaining thousands of line items through bulk structured data file uploads faces reconfiguration work with a hard edge, and older structured data file versions sunset on January 28, 2027, which compresses the schedule further.

The second is that this continues a direction rather than starting one. Google removed digital content label exclusions from YouTube campaigns in September 2024, and the guidance since has pushed buyers toward inventory modes and content themes instead of explicit exclusion lists. Inventory modes are preset bundles. Content themes are positive selections. Neither reproduces what an exclusion list does, which is to state a prohibition in terms the buyer chooses and can audit line by line. A preset expresses the platform's judgement of what belongs together; an exclusion expresses the advertiser's judgement of what does not.

A third change lands on October 12. YouTube responsive advertisement uploads will require a business name of at most twenty-five characters and a logo of at least 144 by 144 pixels at a one-to-one ratio, where the parent advertiser has no default values set. That is a small asset requirement with a large operational tail for any account that has never populated advertiser-level defaults, and it arrives eleven days after the exclusion controls go.

Placed beside the AdMob outage, the pairing is not about blame. It is about the same asymmetry appearing twice in three days. In the first case a rendering path failed and no publisher-side setting could reach it. In the second a targeting path is being withdrawn and no advertiser-side setting will replace it. Both are consequences of running a business on infrastructure whose controls are held one layer above the person accountable for the outcome.

Criteo loses the executive attached to every 2026 announcement

Todd Parsons is leaving Criteo after six years as chief product officer and president of performance media, to become chief executive of a private company the announcement did not name. PPC Land reported on August 29 that Diarmuid Gill, Criteo's chief technology officer, will oversee the product organisation on an interim basis, supported by segment-level product leaders across retail media and performance media. Parsons stays in an advisory capacity from September 1 to September 30, 2026. No permanent successor has been named and no timeline for the search was given.

Criteo's statement was the standard one: "Our product strategy, roadmap and client commitments are unchanged."

Roadmaps rarely change on the day a departure is announced. What changes is who defends them internally when priorities collide, and that matters more than usual here, because Parsons was attached to every significant product initiative Criteo announced in 2026: the Agentic Commerce Recommendation Service, the ChatGPT partnership with OpenAI, the expansion of the GO self-service platform, and the curation deal with TripleLift. Each of those is mid-rollout. Each now sits under an interim structure led by a chief technology officer holding two jobs.

The timing is the story's weight. Executive departures happen; three changes in four weeks during an unresolved takeover approach is a pattern.

On August 5, Criteo cut its full-year guidance as second-quarter revenue fell 11 percent to 428 million dollars, forecasting a 10 to 12 percent decline in contribution excluding traffic acquisition costs. Retail media, the segment the company has spent years positioning as its growth engine, fell 21 percent year on year to 47.9 million dollars. That is a difficult number to reconcile with a category that IAB Europe measured growing 21.1 percent to 13.7 billion euros, and it points to share loss rather than category weakness.

On August 10, Connor McGogney became chief financial officer. On August 29 the product chief left. Sitting behind all three is the approach reported on July 6, when Vista Equity Partners and Quinti Capital were said to have offered a premium exceeding 50 percent for the company. That approach remains unresolved nearly two months later. Criteo also plans to complete a domicile transfer from Luxembourg to the United States in January 2027, having already dropped its French domicile with a New York listing in view.

An acquirer weighing a bid reads a product leadership vacancy in one of two ways. It is either a discount, because the organisation is destabilised and the price should reflect it, or it is a saving, because a private equity owner intends to install its own leadership regardless and no longer has to buy out an incumbent. Neither reading is favourable to public shareholders, and PPC Land has argued that the bid itself is a test of private equity's appetite for a sector whose public valuations have detached from its revenue.

There is a counterpoint in the same week's hiring news, and it runs the other direction. Adweek reported on August 28 that Kim Sizemore, an eighteen-year veteran of Wieden and Kennedy, has joined Finn Partners to build its first integrated media practice. Sizemore will oversee media strategy across a roster that includes Jack Daniel's, 2K, Mazda and Delta, with a brief covering social, in-person activations and traditional advertising. Her description of contextual placement was more memorable than most: her husband, she told Adweek, says his wife is the reason tampon advertisements do not run on ESPN.

The contrast is instructive without needing to be laboured. An independent agency is creating a senior media role that did not previously exist. A publicly listed ad tech vendor under a takeover approach is covering one with an interim appointment. Both are decisions about where product and media judgement is going to live over the next eighteen months, and they point in opposite directions.

Amazon turns the returns pile into 60 million listings

Amazon published its Second Chance Deal Days announcement on August 28, 2026, and PPC Land reported the following day on a European secondhand operation now exceeding 2 billion euros in annual sales across 60 million catalogued returned and refurbished items.

The promotional calendar runs in three windows with different mechanics in each. The United States event runs from August 28 to September 7, folded into Labor Day Deals, applying additional discounts on inventory already marked down. Mexico and Brazil run from August 31 to September 6, branded locally as Semana de Segunda Vida and Semana da Segunda Chance, with Brazil setting a 10 percent floor and Mexico capping at 30 percent off selected items. Europe runs from September 1 to September 10 across the United Kingdom, Germany, France, Italy and Spain, with discounts framed as up to 50 percent against the recommended retail price of equivalent new products.

That last framing is the commercially interesting one. A discount expressed against the new-product recommended retail price puts a used item and a new item on the same price axis, on the same search results page, in the same sponsored placement auction. The comparison is not incidental to the merchandising. It is the merchandising.

The supporting figures give the operation scale. European customers saved 380 million euros buying secondhand over the prior year. United States customers saved more than 1.5 billion dollars across more than 100 million products. Amazon has invested over 4 billion dollars in European returns infrastructure, and customers outside promotional windows save an average of 20 percent against new retail pricing in Europe and the United States. Electronics and home and kitchen are the leading categories.

Rory Feldman, Amazon's European director of returns and recommerce, described the grading process directly: "Amazon inspects all customer returns, testing, cleaning and repairing them where necessary." Items graded as perfect condition return to sale as new. Everything else is discounted by condition tier. The company also said it donated or supported its sellers in donating 216 million items worldwide during 2025.

New this year is authenticated pre-owned luxury from Luxclusif in the United Kingdom and Germany, carrying an additional 15 percent discount during the event window. Authentication is the barrier that has kept marketplaces out of resale luxury, and outsourcing it to a specialist is the standard way in.

For advertisers the consequence is inventory. Sixty million additional listings is sixty million additional product detail pages, each carrying sponsored placement slots, each competing in the same auctions as new-product listings from the same brands. A brand bidding on its own category now bids against discounted units of its own previous production runs, sold by the platform that also sells its advertising. The high-return-rate signals that Amazon uses in ranking operate independently of bid or budget, which means the sponsored products path can be shaped by fulfilment performance that no campaign setting touches.

The wider commerce context has been moving in the same direction. Amazon's advertising revenue rose 26 percent to 19.8 billion dollars in the second quarter of 2026. Measured Prime Day advertising spend in June fell 8.8 percent while conversion rose 17.1 percent, a combination that describes fewer, better-qualified impressions. IAB Europe put European retail media at 13.7 billion euros, one fifth of all digital advertising on the continent. And the shopping surface itself is consolidating, with Amazon merging Rufus and Alexa+ into a single assistant serving 350 million trailing-twelve-month users.

Regulation is shaping the European economics too. The European Union removed the 150-euro duty exemption for low-value imports on July 1, 2026, replacing it with a flat 3-euro charge, which raises the landed cost of the cheap cross-border new goods that resale inventory competes against. NIQ research cited in related coverage put online fashion growth at 3 percent while resale gained ground. A returns operation that was a cost centre a decade ago is now a category with its own promotional calendar, its own authentication partners and its own advertising inventory, and it closes with an FBA inbound cutoff on September 16 that sits immediately after the European window.

Also noted

  • August 27: Gap Inc. reported second-quarter net sales of 3.7 billion dollars, down 2 percent, while the Gap brand itself grew 9 percent to 844 million dollars with comparable sales up 10 percent, a split attributed to culture-led campaigns including the Hailey Bieber partnership and The Hailey Jean product line. Adweek
  • August 28: Tariq Hassan, who stepped down as McDonald's chief marketing and customer experience officer in January 2025, was named chief marketing and customer growth officer at Wendy's, while Lisa McKnight left Mattel to run SC Johnson's lifestyle brands division covering Method and Mrs. Meyer's. Adweek
  • August 28: Target removed a children's Halloween costume after shoppers said its black mesh mask and snaggletoothed grin evoked minstrel caricature, the retailer's second costume withdrawal after a 2020 item that recalled Anne Frank, with Nataly Kelly of Zappi noting that nearly 70 percent of the 2026 Halloween assortment is new merchandise. Adweek
  • August 28: Adweek reported that a Warner Bros. Discovery signal failure took CNN, CNN International and HBO off the air for more than thirty minutes during the noon Eastern hour on August 24, with Dana Bash returning to say there had been a technical issue across the group's channels. Adweek