The open programmatic auction — the sprawling, opaque, machine-mediated market for display advertising that runs against the entire ad-serving open web — has been declared dead more often, by more people, in more publications, than any other single mechanism in the marketing stack. The declarations have been well-supported. Between roughly 2018 and 2024, the share of programmatic display spend flowing through the open exchange fell steadily, from something like 62% to something like 27% of total programmatic activity. Private marketplaces (PMPs), programmatic guaranteed deals, and, increasingly, retail media platforms absorbed the demand. The open exchange kept the long tail of low-quality supply, the arbitrage operators, and — famously — a substantial share of made-for-advertising websites whose only reason to exist was to sell display impressions to inattentive buyers.
Nothing in this story is wrong. What is wrong is the conclusion the industry drew from it, which was that the open exchange would continue to shrink until it was, effectively, gone. Because in late 2025 and the first half of 2026, that shrinking stopped. On the buyer-side data we have access to across 51 audited accounts, open exchange share stabilised in Q4 2025 and, in Q1 and Q2 2026, actively grew back to something like 41% of programmatic activity. Not overnight. Not universally. But visibly, and durably enough that the trend is worth naming.
This essay is about what happened and, more importantly, about why the return is producing a different shape of open exchange than the one that was, quite rightly, being written off five years ago.
What changed
Three things, in order of impact.
The first, and by some margin the most consequential, was the collective repricing of made-for-advertising sites across the buyer ecosystem. Through 2024 and into 2025, several major DSPs, several agency holding companies, and — under quiet pressure from advertisers — a substantial share of independent buyers implemented systematic MFA-detection and MFA-exclusion. The Association of National Advertisers' late-2023 report on MFA spend was a catalyst; the technical infrastructure to act on it was already in place. By mid-2025 the MFA share of open exchange spend had fallen by more than a third across the accounts we audit. This did two things simultaneously: it improved the average quality of what an open-exchange bid was competing for, and it removed a substantial share of supply-side arbitrage that had been distorting the auction's price signals.
The second, and the one the industry writes about the least, was the substantial evolution of curated supply structures — curated marketplaces, curated deal packages, publisher-federated deals — that sit within the open exchange but restrict the supply pool to inventory that has been reviewed at the publisher level. These structures are, technically, still open-exchange auctions; they use the same protocol, the same DSPs, the same bidding logic. What they do is filter the supply pool to a level of quality that the naïve open exchange was not delivering. A buyer who used to only trust PMP structures could, by late 2025, get a substantial share of the same quality through a curated marketplace at meaningfully lower CPMs and with meaningfully better scale.
The third was the maturation of supply-path optimisation as a discipline. Five years ago, most DSPs allowed a buyer to bid on the same impression multiple times across different SSPs, with the winning path determined largely by the network's own auction logic. This was, in aggregate, a substantial source of waste: the same impression, sold through different paths, could be won at different prices, and the buyer was effectively bidding against themselves. Modern SPO tooling now identifies redundant paths, prioritises the most efficient path per supply source, and — critically — surfaces the price differences between paths in ways that make wasteful paths clearly visible to the buyer. Applied at scale across an account, SPO alone recovered somewhere between 8% and 18% of programmatic budget across the accounts we audit. That budget did not, primarily, go elsewhere. It stayed in the open exchange, at higher effective quality.
What the buyers are actually buying
The composition of what a competent buyer is now purchasing in the open exchange is, on our audit sample, substantially different from what they were purchasing four years ago.
Approximately 34% of open-exchange spend, in our sample, now flows through curated marketplaces or curated deal packages — up from perhaps 8% in 2021. The buyer is technically bidding in the open exchange, but the supply pool has been narrowed to publisher-reviewed inventory. The average CPM is higher than the open exchange baseline but meaningfully lower than a comparable PMP would command. The performance, on the accounts we audit, is roughly comparable to PMP spend on a like-for-like basis.
Approximately 27% of open-exchange spend flows against SSP-side inventory-quality filters — SPO-aware bids against publishers that have cleared minimum quality thresholds on the SSP's own metrics (viewability, brand safety, non-MFA). This is a much larger share than the equivalent metric would have shown three years ago, and it reflects the substantial investment SSPs have made in cleaning their own supply pools in response to buyer pressure.
Approximately 22% flows against explicit inclusion lists — the buyer has decided which specific publishers they want to reach and is bidding only against those. This is closer to the traditional media-planning model than to programmatic-as-usual, but it is running through the same open-exchange plumbing.
The remaining 17% flows through the traditional open-exchange model of "bid on any inventory the DSP surfaces that matches audience targeting". This was, five years ago, roughly 80% of open-exchange spend. It is now the minority use case. The buyers who still spend meaningfully here are, on our audits, either doing so deliberately (with sophisticated post-bid quality analysis to compensate for the pre-bid opacity) or doing so out of inertia (which our audits often flag as a source of avoidable waste).
"The open exchange is not the same product it was four years ago. The plumbing is the same; the water flowing through it is meaningfully cleaner. The buyers who wrote off the open exchange during the MFA era are now missing a substantial share of decent supply that has, in the interim, cleaned itself up."
What the walled gardens didn't deliver
To understand why open exchange spend has grown back, it is useful to look at what it grew back from. Between 2018 and 2024, most of the display spend leaving the open exchange did not simply retire; it moved somewhere else. The primary destinations were the walled-garden platforms (Meta, YouTube, TikTok, LinkedIn, etc.) and, more recently, retail media platforms.
The walled gardens delivered, for a while, a substantially better buying experience than the open exchange. They had cleaner inventory, better attribution reporting, more sophisticated audience-targeting, and more consistent delivery. They also had — and this became increasingly consequential — a much higher effective CPM than the open exchange had ever produced. As long as the performance justified the CPM, the trade was worth making. Through 2022 and 2023, on many of the accounts we audit, the performance stopped justifying the CPM. Meta's post-iOS-14 attribution deterioration, YouTube's CPM inflation in the wake of TikTok's rise, and LinkedIn's continued premium pricing all produced accounts where the walled-garden allocation was significantly less efficient than the open exchange allocation had been.
The buyers who noticed this — and, at first, they were a minority — began quietly reallocating. Some of the reallocation went to retail media, which has its own set of problems (a separate essay). Some went to CTV, which is genuinely a growing category. And a meaningful share, on our audits, went back to the open exchange, into the cleaned-up supply structures described above. The buyers who executed this reallocation in 2024 and 2025 are now, on aggregate, running the most efficient programmatic accounts we see. The buyers who did not are, on our audits, spending more per outcome than they need to.
What to do
If you are running a programmatic function right now, three practical implications.
First, run an inventory audit. Pull three months of your programmatic delivery, at the domain level, sorted by spend. Look at the top hundred domains you have delivered against. Are they domains you would defend to your CMO? If a substantial share of them are unfamiliar names, or names you associate with poor content, or names that appear multiple times with slight variations, you probably have MFA exposure worth addressing.
Second, evaluate curated marketplace access. Most major DSPs now offer curated marketplaces through their SSP partners. If you are not currently accessing any, you are almost certainly leaving efficient supply on the table. The onboarding is not trivial but is well within reach of any competent programmatic team.
Third, run an SPO analysis. If you are not currently deduplicating supply paths across your DSP, the recovery potential is, on our sample, somewhere between 8% and 18% of your open-exchange spend. This is not, on any reasonable priority list, work that can wait another quarter.
None of these moves is glamorous. All three are, by our audit sample, high-leverage. The open exchange is, quietly, working again. The buyers who invest in the discipline required to use it well will be, over the next three to five years, disproportionately advantaged over the buyers who continue to treat the exchange as a broken tool.