Here is a fact that should bother anyone who buys media for a living. People spend somewhere between 55 and 60 percent of their online time on the open web, the sprawl of news sites, blogs, forums, weather apps and games that no single company owns. Advertisers spend only 20 to 30 percent of their budgets there. The other 70 to 80 percent flows into a handful of closed platforms where people spend the minority of their attention.
Redseer, the research firm that put hard numbers on this, calls it the internet's most expensive mismatch. Global ad spend crossed one trillion dollars in 2025. A gap of thirty-odd points between where attention sits and where money goes is, by any reading, a multi-hundred-billion-dollar misallocation. And it is not a temporary wobble. The gap has held roughly steady for five years, and the share heading into the closed platforms keeps climbing. This post is about why that happens, what the closed platforms genuinely do better, what the open web still does well, and what it means that the open web is shrinking as a place to spend money.
What a walled garden actually is
The concept is older than the phrase. Back in the 1970s the Bell System leased phones to customers rather than selling them, so the hardware on your desk stayed locked to one company's network. The term itself is usually credited to John Malone, the cable mogul who ran Tele-Communications Inc., who used it in the mid-1990s for a closed platform where the operator controls the content and the access. AOL ran the consumer version of that idea: a curated zone of sponsored content where the company decided what you saw and sold the way in. The web that followed was supposed to be the opposite. Anyone could publish, anyone could link, and an ad could be bought against any page through an open auction.
A walled garden in advertising is a platform that keeps three things inside its own walls: the inventory, the audience data, and the measurement. You buy Meta ads inside Meta's interface, against Meta's audiences, and you measure the result with Meta's own reporting. You cannot easily carry that audience out, and you cannot independently verify that the numbers are true. Google, Meta and Amazon are the obvious three. But the category has grown. TikTok is one. Retail media networks run by Walmart, Target and Kroger are walled gardens built on shopper data. The big streaming platforms, Netflix, Amazon's Prime Video, Disney, increasingly behave like them too. The open web is everything else: independent publishers and apps whose inventory is bought programmatically through demand-side platforms, exchanges and supply-side platforms, with no single owner of the pipe.
The distinction matters because the two worlds run on different physics. Inside a garden, the platform sets the rules of the auction, knows the user, and grades its own homework. On the open web, the impression is auctioned across many independent systems, the user is harder to identify since third-party cookies decayed, and measurement comes from outside vendors. One of these is far easier to buy. That, more than anything, explains the mismatch.
Why the gardens keep winning
It is tempting to frame this as advertisers being foolish or lazy. They are not. The gardens win for reasons that are entirely rational at the level of a single campaign.
Start with identity. A logged-in platform knows who you are across devices because you signed in. That first-party login is an identification layer no independent network can match, and it became far more valuable once Safari and Firefox blocked third-party cookies and Apple's app tracking rules cut mobile signal. The open web spent years bracing for Google to kill the cookie in Chrome too, then watched Google abandon the plan and keep cookies in place by April 2025. Either way the gardens never depended on the cookie. Their addressability was never at risk.
Then measurement. A garden runs closed-loop attribution: it sees the ad impression and, because the user is logged in, it often sees the outcome too, a purchase, an install, a store visit. Retail media takes this furthest. Walmart Connect can tie an ad served on Walmart.com to a sale rung up at a Walmart register. That is a clean, same-platform line from spend to revenue. On the open web, an advertiser stitches that story together from multiple vendors with conflicting attribution windows, and is never quite sure it reconciles.
Then simplicity. A garden offers one buying interface, one relationship, one invoice, one set of automated tools that optimise toward your goal. The open-web programmatic stack is the opposite: demand-side platforms, supply-side platforms, exchanges, data providers, verification vendors, each with a fee and a glossary. Industry audits keep finding that a large share of an open-web dollar disappears into that chain before it reaches a viewable impression. Faced with a tidy garden and a leaky maze, a time-poor team picks the garden. Every time.
The result compounds. eMarketer projects that US walled garden programmatic display spending will keep pulling away from the open web, with the open web's share of programmatic display drifting down toward roughly 19 percent by 2027. Statista's global forecast puts walled gardens above 80 percent of digital ad revenue and still rising. The two fastest-growing channels in all of advertising sit firmly inside the walls. US retail media spending is on track for about 71 billion dollars in 2026, now a peer of search and social. Connected TV is heading toward roughly 38 billion. Streaming passed cable and broadcast combined in US viewing share in May 2025. The money is not just sitting in the gardens. It is growing fastest exactly where the gardens are strongest.
What the open web still does well
This is where the honest version of the story diverges from the gardens' marketing. The mismatch is real, but it does not mean the open web is a bad buy. It means the open web is an underbought one.
First, attention is not the same as zoned-out time. The Trade Desk's editorial arm has made the case that social platforms hold a far smaller slice of consumer attention than buying patterns assume, and that people reach for the open web in a more deliberate, learning frame of mind than the one they bring to a feed. The Trade Desk has a stake in that argument, so weigh it accordingly. But the underlying point is sound: a minute of considered attention on a recipe site or a review article is not worth the same as a minute of thumb-flicking past a feed. The gardens count the time. They do not always own the mindset.
Second, incremental reach. If a brand already saturates Meta and Google, the next dollar inside those gardens mostly hits people it has already reached. The open web is where the rest of the audience is. It is the only practical way to reach a person across many sites and contexts without a platform login as the price of entry.
Third, contextual relevance. The decay of cross-site identity revived an older idea: target the page, not the person. Modern contextual targeting reads what an article is actually about and places an ad to match. It needs no personal identifier, sidesteps the privacy problem, and tends to land where a reader is already engaged with a topic. That capability lives almost entirely on the open web.
Fourth, transparency and independence. Inside a garden you trust the platform's own report. On the open web you can, in principle, see where an ad ran, choose the supply path, and bring in third-party verification and measurement that does not work for the seller. That independence is precisely what a garden cannot offer, because the garden is the seller.
The sharper practitioners have stopped treating this as a binary. The useful question is not open web or walled garden. It is what each is for: gardens for efficient, measurable lower-funnel performance against known audiences, the open web for incremental reach, brand building, contextual relevance and a check on the gardens' self-scored numbers.
What a shrinking open web means
If the open web were merely smaller, that would be a budgeting footnote. The problem is what a shrinking open web does to the things that depend on it.
It depends on publishers. The open web is funded, in large part, by advertising that flows to independent journalism, niche enthusiast sites and the long tail of useful pages. As programmatic dollars concentrate in the gardens, the open auction is sliding toward a marginal share of many publishers' revenue. A publisher that cannot fund itself through ads has three options: a paywall, a deal to feed its content into a larger platform, or closure. Each of those quietly makes the open web smaller and less open. The supply that remains is increasingly polluted by made-for-advertising sites and AI-generated filler built to catch programmatic spend rather than readers, which makes buyers trust open inventory even less, which pushes more money back behind the walls. It is a loop that feeds itself.
It depends on competition. The clearest external check on garden power is the antitrust system. A US court ruled in April 2025 that Google illegally monopolised the publisher ad server and ad exchange markets that the open web runs on. The remedies phase argued through late 2025 over whether Google should be forced to sell those assets or merely change its conduct. Whatever a judge decides, and observers expect a long appeals process after, the case is a reminder that the open web's plumbing is not a neutral commons. It is largely owned by one of the gardens it competes with.
And it shows up in the market's mood. The Trade Desk built a large business as the standard-bearer for open-web programmatic. In 2025 and into 2026 its growth slowed and its stock fell sharply, dropping about 13 percent in a single day after a first-quarter earnings miss, as Amazon undercut open-web demand-side fees and investors questioned the durability of the open-web story. Buyers still call it the king of the open web. The market's worry is plain in the question that title now invites: how big does that kingdom stay.
What to do about it
None of this argues for abandoning the gardens. They work, they are measurable, and for lower-funnel performance against audiences you can already identify they are often the right call. The argument is narrower and more practical.
Treat the mismatch as information. If a competitor pours everything into the same gardens, the open web is where attention is comparatively underpriced. Audit your own split. If your budget mirrors the industry's 70 to 30 tilt while your customers spend most of their time on the open web, the gap is not prudence. It is an unexamined default.
Insist on honest measurement. The gardens grade their own homework, so judge them with something they do not control. Incrementality tests and marketing mix modelling, which look at aggregate outcomes rather than platform-reported conversions, are the tools that reveal whether garden spend is causing sales or simply taking credit for them. They frequently show that the last incremental dollar inside a saturated garden does less than the dashboard claims.
Build the asset the gardens make you rent. Inside a walled garden you borrow the audience and give it back when the campaign ends. Your own first-party data, consented and well structured, is the one identity asset that travels with you, works on the open web, and does not depend on any single platform's goodwill. That is the agentic angle worth taking seriously. As AI systems begin to plan and buy media, and as AI agents start to shop on behalf of people, the advantage shifts to organisations whose customer data and product information are clean, structured and portable rather than locked inside a platform they do not control. Building that foundation now, the work Perform Digital does with enterprise teams, is what keeps a brand addressable wherever attention moves next.
The open web is not finished. But it is being underfunded into a worse version of itself, and the advertisers who notice the mismatch before it fully closes are the ones who will buy attention while it is still cheap.
Council summary
This post argues that the gap between where people spend attention (the open web) and where advertisers spend money (the walled gardens) is real, structural, and widening, and that the gardens win on identity, closed-loop measurement and buying simplicity rather than on superior attention. We verified the load-bearing figures: Redseer's 55 to 60 percent attention versus 70 to 80 percent spend split, the trillion-dollar 2025 ad market, eMarketer's roughly 19 percent open-web programmatic display share by 2027, US retail media near 71 billion dollars and CTV near 38 billion in 2026, Nielsen's May 2025 streaming milestone, the Google ad-tech monopoly ruling, and The Trade Desk's roughly 13 percent single-day drop after its first-quarter 2026 miss. We corrected one origin error, the "walled garden" term was coined by John Malone in the mid-1990s, not by a 1970s telecom executive, and we replaced a mis-attributed cookie source and softened a Trade Desk research claim to reflect its source's bias. The takeaway for a media buyer: treat the mismatch as a pricing signal, judge garden spend with incrementality tests the gardens do not control, and build portable first-party data before AI-driven buying makes that asset decisive.
Comments