Not an SSP
Kraftisms #24 - Two sentences on stage that reframed the whole supply chain debate
Quote(s) of the Week
“If you don’t know where you are going, you’ll end up someplace else.” -Yogi Berra
“A thing is what it does.” -Aristotle
Kraftism of the Week: The Line That Changed the Issue
I had planned to write about events this week, inspired by Marketecture Live (put together by Ari Paparo and Jeremy Bloom), one of the best industry conferences I’ve been to in years. The content was sharp, the vibe was right, and I’ll get to all of that in a future issue.
But something happened on stage that changed what I wanted to write about.
I was sitting in a room with roughly a thousand people - and yes, with a healthy chunk of #thesame300, the same faces I’ve been seeing at these things for twenty years. Mike O’Sullivan, who I worked with at AppNexus, now VP of Product at The Trade Desk, was being interviewed by Chris Kane of Jounce Media. The topic was OpenPath - TTD’s direct integration with publishers that lets their bids reach a publisher without going through an SSP first. It’s been around for a few years and now represents 5% or more of monetization for some publishers.
The question was simple: what is OpenPath? Mike’s answer was the thing that clicked.
OpenPath, he said, is not an exchange and it’s not an SSP. Why? Because it doesn’t try to optimize yield, and it doesn’t broadcast the auction to others.
That’s it. That’s the whole thing. Two sentences, and suddenly a debate that’s been muddying the industry for years came into focus for me. Mike and I worked together at AppNexus, where the supply chain conversation was already complicated enough. Hearing him articulate this so cleanly - on stage, in two sentences - was one of those moments where you think: I’ve been circling that idea for years and just never said it that simply.
Applying the Kraftism: Different Roads, Same Destination
An SSP - supply-side platform - exists to do one fundamental job in the programmatic advertising world: maximize yield for the publishers it represents. It receives bid requests, runs them against multiple demand sources simultaneously, and works to drive the highest possible price for each impression. That’s what makes it an SSP. Not the pipes. Not the technology. The job.
What Mike was saying is that OpenPath isn’t trying to do that job. TTD isn’t trying to optimize yield on behalf of publishers. They’re trying to make sure their bid at least gets seen by the publisher - without a gatekeeper deciding whether it wins the price competition or not. That’s a meaningfully different thing. Publishers can still use SSPs to optimize yield (on the other 95% of monetization). OpenPath doesn’t replace that. It sits alongside it, as a direct line from a specific, significant buyer.
This isn’t entirely new territory. Back in 2017, Criteo went to publishers with a similar pitch: implement our adapter directly, as eventually we’ll shift spend to it from the exchanges. Many publishers tried it. Criteo never fully followed through. They kept buying through the exchanges, too - largely because the industry pushed back on the adapter, and because the math on cutting SSPs out entirely never fully worked. Sound familiar? OpenPath is seeing some of the same pushback today. And Mike was clear on stage that TTD isn’t saying they will stop buying through other SSP channels. They’re running both paths. History doesn’t repeat, but it does rhyme.
So if OpenPath isn’t an SSP, what does that mean for the SSPs themselves?
Quite a bit, actually.
Different Paths. Same Destination.
While the OpenPath conversation has been happening throughout the industry, the SSPs have been making many of their own moves in search of better publisher yield.
PubMatic has been leaning hard into agentic buying. Rajeev Goel, Co-Founder and CEO of PubMatic, has been building toward something meaningful here: infrastructure that lets agents build on top of what PubMatic already has underneath - all the rich systems for packaging, pricing, avails, and campaign setup, and activation. The goal isn’t to wait and see how agentic plays out. They’re iterating, finding new ways for it to add value, and both using and going beyond what existing industry protocols like AdCP currently specify. His view on why this matters is direct: agentic AI is “much more than a technical revolution. It’s also a revolution of the value chain in our industry. It will bring publishers and buyers much closer together.”
Index Exchange is focused on something a little further down in the stack - and probably less visible to anyone not living in the infrastructure layer. Andrew Casale, Index’s CEO, has been talking about a fundamental architectural shift: moving intelligence out of downstream systems and up to the impression itself. The idea is that SSPs sit at the center of real-time decisioning across the open internet - processing tens to hundreds of millions of queries per second - and that’s a unique position to bring AI to bear in a way that isn’t possible anywhere else in the chain. As Casale put it to me: if common calculations that are currently done redundantly by every participant can instead be performed once, at internet scale, techniques that were historically too expensive become practical. That’s not a product pitch. That’s infrastructure thinking.
Magnite is about inventory and data - matching its deep roots across display and CTV with data activation using publisher first-party data, combining that with buyer first-party data and third-party data. The argument is simple. Publishers know more about their audiences at the moment of the impression than anyone else in the chain. Activating that knowledge at the supply layer, layered into deals and inventory packages for buyers, is where Magnite is focused. CPO Adam Soroca put it plainly when I asked him: focus on SSPs that represent supply and have all the data. Advantageous position.
All of these are attempts to make it easier to buy, faster to buy, better to buy, and all that leads to better yield for publishers.
Think of it like Stitch Fix. A stylist - working from data, from signals, from everything they know about what works - curates the best options and sends them your way. That’s the SSP. It’s optimizing across all available demand, assembling the best result it can from everything in the market. But Stitch Fix also lets you shop directly on the site. You see something you want, you buy it. That’s OpenPath. The buyer already knows what inventory they want. They’re not asking the stylist to find it for them. They’re going direct.
Both options exist inside the same system. The stylist doesn’t disappear because you shopped direct once. And shopping direct doesn’t mean you stop trusting the stylist to find things you wouldn’t have found yourself. The curated fixes are still the vast majority of what gets bought.
I’m not taking sides here. I see real value in both, side by side. But the choice of whether to integrate OpenPath, lean on SSPs, or run both in parallel - that’s a decision each publisher has to make for themselves, based on their relationships, their inventory, and who their most important buyers are. What I am saying is that framing this as either/or misses the point entirely. The SSP’s job is yield. OpenPath’s job is access. Those aren’t the same job.
To Wrap It All Up
I’ve been doing my Listening Tour for over two years now, averaging five meetings every week across the whole industry. One thing that comes up more than almost anything: the supply chain conversation, dressed up in different clothes every few months, cycling through the same fundamental tension between what buyers want and what publishers need.
Mike’s two sentences cut through all of it. OpenPath isn’t an SSP because it doesn’t optimize yield. That’s not a criticism of OpenPath. It’s actually the clearest description of what OpenPath is, and what it isn’t, that I’ve heard.
The SSP’s job hasn’t changed. The roads to doing it well are multiplying. The destination is the same.
Andrew
Brilliantly written Andrew. Thank you.
Many years ago I sat with Frank Addante over lunch and he laid out his early singular yield vision for launching the rubicon project (magnite) vs. typical Adnetworks. It wasn’t called an SSP then. Frank believed in singular yield management for a small fee, I was impressed but my feedback was that this will lead to many copycats with fragmented demand which will lead to a loss of purpose because if there’s many than there’s not 1 complete source of truth. That came true, but it took much longer than I anticipated.
So the original purpose of SSPs was singular yield management so one pub ad system could assess true yield and optimize.
Back then, my company Oridian, was the first seat holder on right media, pubmatic and others, and I can attest that this is what lead right media to call the system “yield manager”. The race was on to be the one system of yield record.
Unfortunately this is NOT how it played out. We have dozens of SSPs, and prebid+GAM became the true final system of Yield Record. The SSPs ultimately became merely meta demand sources with unique traffic shapers, and access to preferred dealid paths (which rarely work).
SSPs do manage yield, but for primarily for themselves, not the publisher. This is not a criticism, it’s the evolution of loosing the central mission of a singular control of publisher yield. OpenPath is merely one important demand source into that true publisher Yield Manager which is GAM+Prebid.
At Infolinks I decided to go in a completely different direction because I was a witness to all of these evolutions and did not want to be a victim of them. I decided to avoid prebid and only use JS on the publisher to create and sell the quality select ad placements. We desired to be our own true unique placements and data source of truth. I believed bidding into a common prebid or many SSPs would leave no distinction and being direct to the pub is invaluable.
I could go on for hours on the history mixed with evolving market economics, it’s truly fascinating.
I'd REALLY like to see a statistical analysis of publisher yield with and without OpenPath. If TTD's bidder is optimizing against OpenPath as a preferred path, (and why wouldn't it?) and publishers are getting less yield on that path than others, why should a publisher do it?
I agree with a generalized assessment that most SSPs haven't done the job of maximizing publisher yield properly. But I also feel very strongly that most DSPs don't in any way reward publishers for providing high quality inventory. There's a really simple example of this - when buying EXACTLY the same inventory over Open Exchange versus a PMP, the open exchange inventory version of that supply is fractions of the cost. Why? I have a lot of hypotheses that I'd love to test, but at the end of the day - only the DSP can actually test that.
It begs the question about all the myriad approaches we've taken over the last few decades, each leading to more and more complexity in the ecosystem. For instance, are we doing ourselves (as an industry) a disservice by enabling header bidding? Is it good or bad? I think it had a purpose to serve, but I'm not convinced it's valuable anymore. But that would require real analysis. And what publishers have teams left to do that kind of analysis?
I'm not 'just' railing here - I think it's time for a more thoughtful reevaluation of the ecosystem. Luckily the various AI / Agentic initiatives taking place right now offer some opportunities for reflection. Because one thing I DO NOT believe is that Open Web publishers have no value. I think we have an opportunity to help them carve out a space for themselves. Not quite sure how yet, but it's been on my mind a lot this year.