APPLE'S RECORD AD QUARTER AND THE TRADE DESK'S GROWTH STALL TELL THE SAME STORY — AND MOST MARKETERS ARE READING IT BACKWARDS

Apple just posted its strongest advertising quarter in history. At the same time, The Trade Desk — the perennial poster child for open-web programmatic — suddenly looked mortal, with growth decelerating just as everyone expected its UID and retail media bets to hit escape velocity. Within hours of the earnings calls, the hot take crystallized across Slack channels and agency war rooms: “The signal is clear. Double down on the walled gardens. The open web is done.”

That conclusion is tidy. It’s also almost perfectly backwards.

What Apple’s record and The Trade Desk’s wobble really tell you is that performance channels on the open web are heading into a brief but powerful advantage window — a period when the majority of your competitors will misinterpret the data and overconcentrate spend inside black boxes, while a minority quietly compound gains in more transparent, more governable environments.

To see why, you have to zoom out from the quarter and look at how we got here. Over the past two decades, performance marketing became the hero of the boardroom precisely because it could transform digital intent into measurable sales and tie every dollar back to business results. As one analysis in AdExchanger’s coverage of performance marketing’s blind spot puts it, the industry’s obsession with provable outcomes created a gravitational pull toward lower-funnel demand capture — the clicks and conversions you can present to the CFO on a single slide.

The problem is that when everyone optimizes against the same hyper-attributable environments, you hit a growth ceiling. Customer acquisition costs climb, conversion rates erode, and incremental gains become harder to find because every brand is hunting the same in-market shoppers in the same few walled gardens. Those platforms can still post monster quarters — Apple just did — even as the underlying performance signal for individual advertisers gets noisier and more commoditized.

That’s the context in which you should read The Trade Desk’s slowdown and the broader anxiety about open-web performance. It’s not that programmatic is structurally broken; it’s that the easy arbitrage of third‑party data and undifferentiated audience buying is finally priced in. Even insiders are pointing out that the old model of paying hefty markups for centralized data — the very dynamic that, as one AdExchanger deep dive into Adobe’s custom algorithms notes, has “tripped up” players reliant on expensive third‑party marketplaces — is losing its edge.

At the same time, the competitive battlefield where performance is actually decided is shifting underfoot. Consumers are increasingly making decisions inside AI‑mediated experiences where the recommendation itself is the ad, whether that’s a retail assistant, an operating-system-level helper, or a conversational shopping bot. As one recent analysis in MarTech’s exploration of AI‑native advertising argues, if your product isn’t surfaced in the synthesized answer, you effectively don’t exist at the moment of intent.

Here’s the catch: those AI‑mediated surfaces are only as good as the data, creative intelligence, and feedback loops you feed them. Walled gardens can abstract all of that into a single “performance easy button,” but they also lock marketers out of seeing how the sausage gets made. You don’t own the algorithm. You don’t really own the learnings. And you’re bidding against every other brand pressing the same button.

On the open web, by contrast, the constraints are turning into a competitive feature. Privacy rules, signal loss, and identity deprecation forced a new generation of infrastructure that connects creative, context, and outcomes in real time. Systems like the creative‑effectiveness loop described in Search Engine Journal’s analysis of DAIVID and ADIN.AI — which score content at scale and link those scores directly to media performance — are a preview of where high‑leverage performance marketing is headed: not a return to the wild west of third‑party cookies, but a governed, data‑rich open web where you can actually see and shape the variables that matter.

That’s the “golden window” hiding in the headlines. While your competitors read Apple’s blowout ad quarter as a mandate to pour even more budget into opaque systems, you can treat it as a timing signal: this is the moment to reallocate attention, not just dollars, toward open‑web performance channels where you still control the levers — data, creative, context, and measurement — that will decide whether your brand is findable in the AI‑first decision journey.

Most marketers will chase the apparent safety of the gardens. The ones who win the next phase of performance will be the ones who read these same earnings reports as a prompt to do the harder, more valuable work in the spaces everyone else is prematurely abandoning.

Apple’s blowout ad quarter and The Trade Desk’s slowdown are being treated like opposing signals: one says “walled gardens are winning,” the other “the open web is dying.” That’s the wrong interpretation. They’re both symptoms of the same structural shift — a move away from undifferentiated, volume‑driven impressions toward environments where attention, identity and context are tightly controlled.

Start with what’s actually happening on the open web. The easy narrative is “AI stole our clicks,” and there’s truth there. Generative answer engines have begun to “restructure the web without partaking in the click‑based value exchange”, which guts the classic SEO → article → display ad funnel. But the deeper trend predates AI. Direct, habitual visits to publishers have been eroding for years, especially among under‑35s. Similarweb data cited by Search Engine Journal shows direct traffic down double digits across popular, premium and public‑service publishers, with some mass brands losing more than half their direct audience.

If you’re a demand‑side platform built on the premise that there will always be abundant open‑web inventory and stable cookies to stitch it together, that’s an existential problem. Even if identity alternatives like UID2 work technically, they can’t fully compensate for the fact that habitual, direct visits to broad news and content sites are shrinking while people spend more time in closed environments and utilities — phones, messaging apps, streaming interfaces, commerce platforms.

Marketers, though, often read a flat quarter at The Trade Desk as a referendum on “programmatic” rather than on the underlying supply and signal. The instinctive response is to hunt for the next mass reach channel: maybe that’s retail media, maybe CTV, maybe AI chatbots. You can see this mindset in the breathless attention to OpenAI’s forecast that its ad business could someday hit $100 billion, a target that, as AdExchanger noted, is nearly twenty times larger than eMarketer’s entire 2030 US chatbot ad market projection. The fantasy is that some new surface will simply replace the cookie‑soaked open web and restore the old scale economics.

But look at what’s actually performing and where capital is flowing. Apple isn’t suddenly “good at ads” in the abstract. It controls the operating system, the identity layer, the app store, the maps product and the default surfaces people touch dozens of times a day. That means extremely high signal density (device, location, context) and strict control over ad load in properties like Search Ads in the App Store and sponsored units in News and Stocks. When AdExchanger flagged Apple’s “renewed interest” in growing its ads business, that wasn’t a pivot into random banner inventory; it was a bet that marketers will pay premiums for safe, high‑intent, high‑identity environments even if the raw impression count is smaller.

The same pattern is visible in other corners of digital. In web push advertising, for example, the industry is explicitly moving away from “spray everyone” tactics. RollerAds’ analysis on MarTech describes a shift from a “volume‑driven phase” toward performance and ROI, where low‑quality traffic is squeezed out, overall volume falls, and each subscriber becomes more valuable. In the short term, that looks like volatility and higher unit costs; in the long term, it creates a healthier market with higher engagement and better outcomes.

Taken together, these signals don’t say “open web bad, Apple good.” They say the era of cheap, low‑signal reach is ending across the board. Any platform whose economics depend on infinite, commoditized impressions is going to feel pressure. Any platform that can offer scarce, high‑signal, high‑trust attention — whether that’s an OS, a retail media network, a premium publisher with a resilient direct audience, or a refined push network — is going to command a growing share of budget.

Marketers misread Apple’s record ad quarter and The Trade Desk’s stall when they treat them as proof that they must pick a team: walled gardens versus the open web. The real divide is between environments that can prove relevance and outcomes with first‑party or contextual signal, and those that can’t. Apple is on the right side of that divide. Increasingly, so are the parts of the “open web” willing to get smaller, cleaner and more intentional.

Performance marketing didn’t start in a walled garden, but walled gardens are where its blind spot became truly dangerous.

For two decades, TikTok-ads-payment-problems-how-to-add-a-payment-method" target="_blank" rel="noreferrer noopener">digital advertising has been optimized around a single seductive promise: if you spend $X today, you’ll see $Y in attributable revenue tomorrow. As commerce moved online and pixels spread across every checkout page, marketers learned to track individuals across sites, tie impressions to conversions and argue, with apparent precision, for incremental lift. As one analysis of performance marketing’s evolution put it, the industry’s energy went into “improving targeting, refining attribution models and optimizing performance campaigns to generate stronger returns,” building a machine that could compete ruthlessly for in‑market demand rather than create new demand in the first place, as AdExchanger’s overview of performance marketing noted.

Walled gardens took that logic and cranked it to 11. Platforms like Apple, Google, Meta and Amazon sit on the richest seams of intent and identity: search queries, purchase histories, app behaviors, location trails and device‑level identifiers. They also own the measurement rails. When you buy their performance products — from Google’s Performance Max to Meta’s Advantage+ Shopping — you aren’t just buying media. You’re buying into a closed optimization loop where the platform chooses who sees the ad, tracks what happens and grades its own homework.

That loop feels like a superpower because it collapses friction. Instead of stitching together disparate data sources, you pour budget into a single product, upload creative, toggle a few goals and watch the machine optimize. Adobe’s head of product even described these systems as “performance easy buttons,” a phrase echoed in coverage of how Google’s PMax and Meta’s Advantage+ leverage their own first‑party data to drive results. Marketers see clean dashboards, strong ROAS and a comforting story for the CFO: “Our spend is working. Here are the numbers.”

But that simplicity is precisely where the blind spot forms.

First, the garden controls the lens. If the same platform defines the audience, auctions the impressions and declares the conversions, it can over‑credit its own role in outcomes. Incrementality becomes a faith statement, not a provable fact. As the earlier examination of performance marketing’s “growth ceiling” argued, brands end up “competing for the same pool of in‑market consumers,” driving up acquisition costs and squeezing incremental gain, even while last‑click dashboards still look healthy, as AdExchanger’s blind‑spot analysis pointed out.

Second, the garden hides the levers. PMax works by blending data across YouTube, Search, Gmail and more, but the marketer never sees which surfaces, queries or creative variations are actually driving outcomes. That opacity is by design: if you can’t see under the hood, you’re less likely to reassemble the engine elsewhere. In contrast, Adobe’s custom algorithm product was explicitly positioned as an antidote: it only uses a brand’s first‑party or trusted partner data and runs on inventory Adobe doesn’t own, giving advertisers more visibility into what’s working and why, as coverage of Adobe’s approach emphasized.

Third, the garden makes short‑term performance feel like the whole game. When every report is calibrated around conversions and ROAS, anything that doesn’t show immediate, trackable payoff looks like waste. Upper‑funnel investment collapses into whatever the platform can retrofit into a performance metric, even though brand outcomes — awareness, consideration, trust — unfold over quarters, not days, and are notoriously hard to measure with the same precision. That’s how demand capture quietly eclipses demand creation: the metrics you can see in‑platform starve the ones you can’t.

You can watch this same dynamic play out beyond classic search and social. As web push advertising has evolved, for example, the market is explicitly moving away from raw volume toward “performance- and ROI-oriented environments” where low‑quality traffic declines and each subscriber becomes more valuable over time, as a recent analysis of web push trends described. That’s the open web slowly adopting the same performance mindset that walled gardens industrialized — and running into the same questions about how to value quality, attention and long‑term customer value when your instruments are tuned to clicks and short‑term conversions.

The irony is that nobody set out to build this blind spot. Walled gardens simply made performance marketing’s biggest success — the ability to prove outcomes — so convenient and so self‑contained that it crowded out everything that’s harder to see. Apple’s record ad quarter and The Trade Desk’s stall are both signals that this era of easy, attribution‑driven comfort is ending. The marketers who misread those signals as a simple “walled gardens good, open web bad” story are clinging to the blind spot instead of acknowledging it.

The open web isn’t dying. It’s hardening.

For publishers, that feels like decline. For performance buyers, it’s exactly what they’ve been training for — they just don’t recognize it yet.

Look at what’s actually happening to open‑web supply. Habitual, direct visits to publishers are collapsing. Similarweb data across multiple outlets shows double‑digit drops in direct traffic for popular, premium and even public service publishers, with titles like the Birmingham Mail and The Mirror losing more than half of their habitual audience in three years, while a more subscription‑oriented Telegraph barely budged, as Search Engine Journal’s analysis of publisher traffic makes clear. Under‑35 audiences are disappearing even faster than older cohorts — exactly the group publishers hoped would become tomorrow’s paying subscribers.

At the same time, the traffic they do get is increasingly non‑human. AI bot visits grew 187% year over year while human traffic crept up just 3.1%, according to the same Search Engine Journal data. Large language models behave like voracious “answer engines,” soaking up publisher content to serve user intent without ever paying it back in clicks. The old value exchange — I give you an article, you give me a pageview and maybe an ad impression — is breaking down precisely where it used to be most reliable: habitual, direct relationships.

From a publisher’s perspective, this looks existential. Direct audiences — the people most likely to subscribe, register or consent to tracking — are eroding. Referral and platform traffic are fickle. And as inventory that can be cleanly measured, consented and identity‑rich shrinks, the rest of the open web starts to look like a low‑rent back alley: bots, made‑for‑arbitrage pages, and users arriving through opaque paths who can’t be tied back to stable IDs.

But for performance‑oriented marketers, those same forces are quietly making the open web more interesting, not less.

First, the low‑quality “spray and pray” impressions that used to pad out programmatic plans are finally being forced out. As one analysis of web push advertising notes, enforcement mechanisms and restriction systems are getting better at filtering “low‑quality traffic sources and questionable practices,” pushing the ecosystem toward “more selective but higher quality” inventory where “volume is nothing if it is not relevant and of poor quality,” as the RollerAds team explained in their look at web push market trends. That logic isn’t confined to notifications — it’s the direction of travel across open‑web formats: fewer junk impressions, more pressure on signals that actually correlate with outcomes.

Second, marketers are being forced to abandon the lazy assumption that “more reach + last‑click attribution = growth.” As demand capture has crowded out demand creation, brands have run headlong into a growth ceiling: everyone is bidding against the same in‑market shoppers, and customer acquisition costs keep rising even as conversion rates fall, as AdExchanger’s breakdown of performance marketing’s blind spot points out. Open‑web buying, stripped of its cheap, low‑signal filler, pushes performance teams toward a different game: capture plus creation, and short‑term ROAS plus long‑term customer value.

You can already see how the savvier buyers are adapting. In competitive intelligence data from social and programmatic auctions, the real edge doesn’t come from a single magical platform but from “a broader media buying system operating across channels,” where signals in CPMs, geography and placement mix reveal who’s quietly winning on efficiency before the rest of the market notices, as one analysis of auction‑level data from Polaris AI argues. Translate that to the open web and the pattern is the same: the advantage shifts from whoever can buy the most impressions to whoever can interpret weak signals across fragmented, higher‑quality supply and move first.

The quiet shift, then, is this: the open web is getting less forgiving for publishers who depend on undifferentiated volume and habitual, unstructured traffic. But it’s becoming a richer hunting ground for performance buyers who can live without walled‑garden certainty and are willing to compete on attention quality, contextual relevance and probabilistic identity instead of raw tonnage.

Marketers misread this when they see Apple’s record ad quarter or The Trade Desk’s slower growth and conclude “walled gardens are winning, so the open web must be dead.” What’s really happening is that the easy, low‑signal open web is dying. The hard open web — the one that rewards sophisticated measurement, creative that actually earns attention, and a willingness to invest in future demand — is just getting started.

Inside Apple’s and Meta’s walled gardens, the most interesting thing isn’t how much top advertisers are spending. It’s how they’re behaving in the auction when the stakes are highest.

If you want to understand where the smartest money thinks growth will come from, watch what they do when every impression is scored in real time, every bid is a prediction, and every inefficiency is punished instantly. Closed auctions — Apple Search Ads, Meta, TikTok, Amazon — compress all the second‑guessing and slide‑deck theory into a single, brutally honest output: how aggressively a brand is willing to compete for a given kind of attention, at a given moment, on a given surface.

When AdExchanger’s analysis of Meta auctions looked across major US insurers, the striking point wasn’t just who spent the most. It was that Progressive, as the category’s largest spender, was simultaneously paying dramatically less per thousand impressions than peers. In most auctions, scale and efficiency move in opposite directions; push for more volume and your CPMs drift up. Progressive was breaking that gravity.

That pricing anomaly is a live signal. It tells you that somewhere in Progressive’s acquisition system, the models, creative, audience construction and funnel design are working together so well that the algorithm is rewarding them with cheaper reach. It’s not “brand strength” in the abstract; it’s operational excellence being recognized by a machine that only cares about predicted performance.

Marketers misread that kind of signal all the time.

They look at Progressive’s cost advantage and think, “We should copy their budget level,” instead of asking, “What kind of conversion behavior and post‑click economics must they be feeding back into the auction to earn this?” They see a spike in competition for a high‑intent placement in Apple’s search ads and assume “the channel is too expensive now,” instead of asking, “Which competitors are still bidding aggressively and what does that say about their LTV math that we’re too scared to model?”

Closed auctions are telling you three things about your best competitors:

1. Their conviction about future demand.
   If a sophisticated advertiser is content to let lower‑funnel placements soften while leaning harder into mid‑ and upper‑funnel formats inside Meta or TikTok, they’re implicitly signaling that they believe they’ve hit a demand‑capture ceiling — and need to create new demand. That’s the blind spot performance specialists have walked into: optimizing ruthlessly for today’s conversions while underfunding tomorrow’s buyers.

2. Their belief in their own creative system.
In an environment where creative can be evaluated and evolved continuously, the brands that win auctions at scale are almost always the ones that have built a feedback loop between creative performance and media execution. What DAIVID and ADIN.AI describe as a “live loop” — score creative, link it to media results, re‑allocate budget in real time — is effectively what Meta’s delivery system is already doing for you. If your competitor’s creative is structurally more effective, the platform will “subsidize” their reach by making it cheaper for them to win.

3. Their clarity on customer economics.
High, sustained bids in brutally competitive placements — branded search inside Apple, high‑intent queries in Amazon, aggressive remarketing on Meta — are often a tell that your rival has either superior unit economics or a clearer view of lifetime value. They’re willing to run thinner (or even negative) on first‑touch ROAS because their models, not their dashboards, are steering the ship.

The mistake is thinking these signals only matter inside the walls where you see them.

If Progressive is earning abnormally cheap reach on Meta, that’s not just “a Meta thing.” It means their system for selecting audiences, orchestrating creative, and scoring outcomes is simply better aligned with how modern recommendation engines work. That system doesn’t stop at the edge of a walled garden. It will travel wherever there is an auction that rewards predicted performance — including the “hardened” open web you’re probably under‑investing in.

So how do you exploit what you see in closed auctions outside of them?

1. Translate auction outcomes into hypotheses about your funnel.
If you’re consistently outbid on certain cohorts or placements by the same set of competitors, assume they either:

• Value those users more (different LTV or cross‑sell strategy), or
• Convert those users more efficiently (better onboarding, pricing, or product‑market fit), or
• Tell a story to those users that works better (creative, offers, sequencing).

Each of those is testable outside the garden. You can build open‑web experiments that isolate those variables — different offers, different sequences of storytelling across display and video, different post‑click experiences — and see which combinations narrow the performance gap.

2. Bring AI‑native creative and optimization out into the open.
The same real‑time creative optimization loops that MarTech highlights in AI‑native advertising — hundreds of variants, continuous learning, autonomous budget shifts — can be pointed at programmatic display, native, CTV and publisher‑direct buys. The open web may be getting harder for publishers, but that also means fewer unsophisticated bidders and more room for performance‑driven buyers willing to do the work.

If Unilever needs an infrastructure that can score content from 300,000 creators and connect it to media performance, as Search Engine Journal explains, then your open‑web plan needs something similar in spirit: a system that doesn’t just traffic variations but continuously ranks them, routes impressions toward what’s winning, and retires what’s not — before a quarterly brand study tells you what already failed.

3. Use closed‑auction behavior as your competitive research lab.
You don’t need to match a rival’s spend to learn from their moves. You need to watch:

• Which surfaces they appear to prioritize (short‑form video vs. feed vs. search‑adjacent formats).
• How quickly their creative rotates and iterates.
• When they tighten or relax frequency, relative to seasonality.

Those patterns reveal where they believe incremental attention still exists — and how they value it. You can then go looking for analogous surfaces on the open web: the video placements, contextual environments and publisher partnerships that rhyme with the behavior you’re seeing inside Apple and Meta, but are still mispriced because fewer buyers are applying that level of intent.

Underneath Apple’s record ad quarter and The Trade Desk’s wobble is a single message from the most advanced advertisers in the world: auctions are x‑raying your strategy in real time. The winners aren’t just buying more of the same cheap clicks. They’re teaching algorithms to prefer them — and then exporting that advantage everywhere those algorithms roam.

Most marketers treat “competitive intelligence” as looking at who’s spending the most and what their latest hero video looks like. That’s not intelligence. That’s tourism.

If you want to find under‑crowded open‑web opportunities, you have to read the same signals Apple‑class buyers are reading inside closed auctions — then flip them outward to the rest of the internet.

The raw material is already visible. Platforms like Polaris AI are tracking auction‑level data across social and the open web: CPMs, share of voice, placement mix, geography, creative patterns. The power isn’t in a single metric; it’s in how those signals move together.

When AdExchanger’s analysis of Meta activity showed Progressive simultaneously buying more inventory and paying dramatically less per impression than competitors, the point wasn’t “Progressive spends a lot on Meta.” The signal was structural: a media system built on audience precision, diversified placements and ruthless efficiency. That’s exactly the kind of system that becomes even more advantaged when you move from expensive, crowded walled gardens to cheaper, overlooked corners of the open web.

You can reverse‑engineer those systems to find where the crowd hasn’t arrived yet.

Step 1: Translate auction signals into hypotheses, not headlines

Instead of asking “Who’s spending the most on Meta this quarter?”, ask questions competitive tools can actually answer:

• Where is a category leader’s CPM falling faster than the category average?
• Which placements did they ramp up right before that efficiency gain appeared?
• In which geos or audience segments do they suddenly tolerate higher CPMs?

As AdExchanger’s breakdown notes, these micro‑shifts in allocation and efficiency are where strategic advantage shows up first. If a competitor’s feed CPM plateaus but Reels or Stories CPMs keep dropping while spend explodes, you’re watching them arbitrage an underpriced format in real time.

Your job isn’t to copy the exact tactic inside Meta. It’s to form a hypothesis: “This advertiser is systematically overweighting short‑form, high‑attention placements that the rest of the market hasn’t priced correctly yet.” That thesis travels.

Step 2: Map those theses to open‑web supply

Once you know how a sophisticated buyer wins inside closed auctions, look for structurally similar surfaces on the open web:

• If they’re leaning into high‑frequency, low‑friction short video, there are likely under‑monetized video units on mid‑tier publishers and vertical networks.
• If they’re concentrating spend in certain regions at higher CPMs, local news, review sites and community forums in those geos may be undervalued relative to the quality of attention they deliver.
• If they’re achieving low CPMs through distributed placements, there are probably open‑web exchanges and SSPs where similar multi‑placement strategies are still cheap because brand buyers haven’t fully shown up.

Marketers obsess over whether a given publisher is “brand safe” or “on plan,” but ignore the deeper question: where does the structure of this inventory rhyme with the structures that are winning in closed auctions?

Step 3: Bring AI‑native evaluation with you

The reason top advertisers can pounce on these opportunities is that they don’t guess; they instrument. In the same way DAIVID’s creative effectiveness models are wired directly into ADIN.AI to create a live loop between creative intelligence and media execution, you need a loop that makes the open web governable at speed.

On the open web, that means:

• Scoring creative variants in real time against open‑web performance, not just social metrics.
• Auto‑scaling winners and killing losers across exchanges, placements and publishers as fast as the auction data changes.
• Folding historical performance back into planning so you can spot patterns in which environments consistently over‑deliver.

As Search Engine Journal’s coverage of Unilever’s AI‑driven creator network points out, once content scales into hundreds of markets and platforms, human‑only evaluation collapses. The same is true when you deliberately spread spend across dozens of open‑web pockets the rest of your category is ignoring. Without AI‑native evaluation, you won’t see the arbitrage until it’s gone.

Step 4: Treat the auction as your early‑warning radar

Most “competitive” decks are quarterly postmortems. By the time your team notices a rival shifting TV budget into CTV or social, the price has already risen. What matters now is what MarTech describes as autonomous optimization: systems that pick up weak signals and move budget before your media calendar even updates.

If your intelligence stack is watching the same real‑time allocation and efficiency shifts that Apple‑class buyers are acting on inside their own walls, you gain something incredibly valuable on the open web: a few weeks’ head start. That’s how long it usually takes for the herd to notice a bargain and bid it away.

The under‑crowded parts of the open web aren’t hiding. They’re lit up in your competitors’ auction behavior every day. You just have to stop treating that behavior as a scoreboard — and start treating it as a map.