The Trade Desk launched OpenPath back in February 2022 as a direct-to-publisher supply route, basically letting publishers plug into TTD without an SSP sitting in the middle, and TTD turned off Google Open Bidding the same day the product launched. The pitch was simple enough: fewer middlemen, more of the media dollar actually reaching the publisher, a cleaner price for whoever's buying.

It was controversial pretty much from day one, though not for the reason most buyers think about now. Publishers were worried OpenPath would pull money out of the market, and SSPs, understandably, suspected it would squeeze their margins at best and cut them out entirely at worst. One SSP exec put it bluntly at the time: "SSPs have to be more than just a pipe." That was the 2022 fight, and it was mostly about who gets cut out of the chain.

The 2026 version of this fight is a different animal though, less about who's in the chain and more about whether buyers actually know what they're paying for in the first place. Back in February, Adweek reported that Dentsu and WPP had quietly pulled back from OpenPath over hidden fees and a lack of transparency around where ads were actually running. TTD pushed back hard on this, its CMO responding directly: "There are no hidden fees beyond that... if OpenPath is selected by an advertiser, it's simply because it's the cleanest, most cost-efficient path."

Then the following month it got bigger, Publicis told clients in a memo that later leaked that an independent audit found The Trade Desk had improperly applied its DSP fee on top of other fees, and had billed clients for tools they'd been auto-opted into without any proof of authorization. TTD's stock dropped more than 7.5% on the news, and Omnicom said it would run its own audit too. Rival DSPs started circling pretty much immediately, StackAdapt and Quantcast both reaching out to agency buyers directly to pitch a switch.

Start with the number OpenPath gets sold on. Across the full portfolio here, OpenPath impressions came in about 43% cheaper on CPM than the same inventory bought through named SSPs.

It holds up structurally too, though unevenly by placement and device. Desktop OpenPath CPM ran 51% below non-OpenPath desktop CPM, the biggest discount of any segment, and below-the-fold inventory showed an even steeper 72% discount. The one exception worth flagging is CTV, where OpenPath CPM actually came in 9% higher than non-OpenPath, on top of already being the lowest-volume, least reliable segment in this dataset.

So on the metric everyone quotes, OpenPath delivers. The question that actually matters, though, is whether it's priced correctly, whether you're paying less for a genuinely comparable impression, or paying less because the impression itself is worth less. CPM alone can't answer that, so I went and checked the actual attention environment next.

Across the portfolio, OpenPath's CPA ran 31% higher than non-OpenPath, and ROAS came in 28% lower.

Breaking it out by fold position and device tells you this isn't a placement-mix artifact. Above-the-fold OpenPath inventory still ran 24% worse on CPA and 24% worse on ROAS than above-the-fold non-OpenPath - the segment where OpenPath should theoretically look most like named-SSP supply. Below-the-fold was worse still, 43% worse on CPA. On desktop, where the CPM discount is deepest, CPA still ran 30% higher. If this were about where or how the ad renders, controlling for fold and device would close the gap - it doesn't.

I checked this against a second, non-overlapping dataset from earlier this year covering the same portfolio, and the CPA and ROAS gaps held in the same direction and a similar range.

I ran a separate analysis, using impression-level attention data from a different TTD reporting stream, specifically to test that. Four metrics, portfolio-wide, OpenPath versus non-OpenPath:

Three of four metrics land neutral to slightly favorable for OpenPath, and the full distribution backs it up. OpenPath actually has more impressions landing in the best-quality refresh bucket and more impressions in the lowest-competition viewport bucket than non-OpenPath does. By every attention signal in this dataset, you are getting the same environment on OpenPath that you're paying 43% more for elsewhere.

That's the good news, and it's a real result worth sitting with, because the discount doesn't look like a quality markdown so much as the market correcting a price that named-SSP inventory may have been charging without the environment underneath it to actually justify.

It's also where the story stops being simple. On this same dataset, OpenPath still ran a 15% lower CTR, a 4.5% higher CPA and a 6.5% lower ROAS than non-OpenPath, so you've got a comparable environment producing a worse outcome, and I don't have a confirmed explanation for that gap, and I'd rather say that plainly than reach for one that sounds good.

One hypothesis worth testing, though: OpenPath routes around the SSP layer, and SSPs also carry a lot of the audience and identity infrastructure a DSP leans on to match the right person to an impression, things like contextual signals, first-party audience extensions, enriched bid stream data. If that layer is thinner on OpenPath, you could end up with the same page, the same viewability, even the same person, and still convert less, simply because the audience match itself is weaker going in. That's a plausible read and not a proven one, so call it TBD until someone actually runs the match-rate data and checks.

OpenPath is doing a genuinely good job on price. It's finding impressions that cost meaningfully less while delivering the same attention environment as the pricier named-SSP alternative, and the gap between the two is big enough that if you're running little or no volume through OpenPath right now, there's a decent chance you're overpaying elsewhere for nothing, which is worth acting on regardless of anything else in this issue.

The platform performance numbers tell a different story though, and I'm not going to wave that away just because the pricing story is good news. CPA and ROAS both come in worse on OpenPath, consistently enough across the portfolio that it doesn't look like noise.

Before you treat that gap as the final word on whether OpenPath actually underperforms, there's something worth sitting with, which is that these are last-touch attributed conversions including post-view credit, and that's a fairly blunt way to measure impact since last-touch and post-view crediting reward whichever impression happened to sit closest to the conversion rather than telling you whether that impression caused anything. There are a few different ways that could be playing out here rather than just one. If OpenPath's impressions tend to land earlier in someone's path, an attribution model built around last-touch is always going to undercount them relative to whatever's sitting closer to the conversion, regardless of which one actually did more work. Identity resolution is worth sitting with too, since OpenPath's direct-to-publisher setup may see a different mix of identifiers than the named-SSP path does, and if match rates or identity coverage differ even slightly between the two, some of what looks like a performance gap could really just be a measurement gap, conversions that happened but didn't get stitched back to the impression that drove them. And it's possible the platform's own crediting logic just weights these paths differently in ways that have nothing to do with which one actually influenced the buyer. Put those together and there's a real chance OpenPath isn't underperforming on business outcomes at all, it's just getting credited less for outcomes it's actually contributing to, for reasons that have more to do with how the measurement is built than with the media itself.

I'll be honest, the size of the gap genuinely surprised me, and I went looking for these explanations because the number didn't sit right on its own, not because I had a theory going in and went looking for support for it. And to be clear, none of this is me explaining the gap away. I think it's a real possibility that the attribution and identity story accounts for a meaningful piece of it, but I don't know that yet, and until someone runs an actual incrementality test I'm not going to pretend I do. What I'd personally do in the meantime, especially for anything bottom-of-funnel with a hard conversion goal attached, is treat this as a reason to pause or scale back OpenPath there and investigate further before committing more budget, because the price advantage isn't close to large enough to cover a CPA and ROAS gap this size even in a world where measurement is part of the story. Upper-funnel, where you're not leaning on last-touch credit in the same way, I'd be a lot less worried about any of this.