What triggered the sell-side curation craze?
Curation is fundamentally a reaction to programmatic’s worsening queries-per-second problem. That’s according to Joe Root, CEO of Permutive, a DMP that offers curation services.
There are too many SSPs in the market, Root said. So DSPs throttle bid requests from SSPs to keep their data processing manageable.
“Say a publisher gets a million impressions [per day] and has 10 SSPs seeing exactly the same impressions,” Root told AdExchanger. “That means the DSP is seeing 10 million bid requests, and that’s too much to process. So the DSP cuts it down to two million. Each SSP now only has an allocation of 200,000 bids, so it’s only going to send requests it thinks the DSP will bid on.”
But DSPs are biased toward bidding on impressions that have a third-party cookie or some other identifier attached, Root said. This fixation on IDs means “DSP are bidding on a smaller and smaller pool of inventory,” and therefore “CPMs are going up, CPAs are going up and performance is going down,” he said.
But a deal ID associated with a curated PMP can serve the same purpose as a cookie or another identifier.
“The SSPs kick the deal ID over the DSP,” Root said. “And The Trade Desk always bids on deal IDs, so it’s really smart to do so.”
Root spoke to AdExchanger about the market dynamics behind curation and whether the rise of PMPs spells the end of the open web.
AdExchanger: The IAB Tech Lab says it created a curation framework years ago by introducing Seller Defined Audiences (SDAs) and other features. But publishers say buyers aren’t interested in SDAs as long as third-party cookies are around. Why are SDAs such a hard sell?
JOE ROOT: Publishers made SDAs available to SSPs and other players, and they never got adopted. SDAs should be ingredients for curators to bake the cake. Curation lets you make your SDAs available inside SSPs and stock their shelves with your data.
But to drive adoption, you have to go to market. No one’s going to turn up and buy your audiences. You have to educate agency teams on why they should pick your audiences versus something from a curator or anything else. You need to be calling agencies every week to see how the audiences are performing. Publishers have to treat it as a channel, which they aren’t today.
How should pubs approach that go-to-market effort?
If I was a publisher, I would have three taxonomies: my direct taxonomy, which you only get if you buy direct; my PMP or programmatic guaranteed taxonomy, where I’m also getting buying data back so I can optimize; and then a fairly generic, non-endemic taxonomy. The third tier fits into the curation bucket.
How granular should these taxonomies be?
Direct should have your most granular taxonomy.
In curation, publishers are competing for the agency trader that right now goes into the DSP and loads up Nielsen’s technology enthusiast segment. That gets spent across the entirety of the open web, including on made-for-advertising sites, and only on cookie environments.
Publishers need to be stocking their data into technology enthusiast packages within SSPs, where it’s getting bought on publisher signals and non-cookie inventory. And it drives performance, because right now that inventory is massively underpriced.
Where does the IAB Tech Lab’s content taxonomy fall short?
Look at how the third-party data companies market themselves. They send emails – here are our Black Friday audiences. Here’s my audience for the Jake Paul vs. Mike Tyson fight.
No one’s out there like, “I want to buy IAB category ‘family,’ sub-category …” – they’re just so boring. It’s the least sexy thing in the world.
The IAB has a good technical specification for how to push audiences through OpenRTB. That’s all anyone needs from them. Let the industry do the rest – defining the audiences, evangelizing and marketing them.
Does curation enable supply-path optimization (SPO)?
We need a supply-chain object. Agencies think a deal ID equals SPO, and if I’m buying a deal ID, I’m buying a clean supply. That’s not true.
Most curated audiences have four or five hops, with people taking 20%, 40%, even 60% at each hop. It’s a massive arbitrage game today. Someone needs to make it explicit how many hops are involved in that supply chain and what everyone’s take rate is. Otherwise, SPO is impossible.
You’re going to see a new host of measurement companies come in to shed light on that. SSPs should take control of that situation before it gets out of their hands.
Curation used to happen mainly within DSPs, but it’s shifted to SSPs. Is there a backlash coming where DSPs try to take back control?
If I’m The Trade Desk, do I get rid of most of my supply-path vendors? It’s a hard problem. I think The Trade Desk CEO Jeff Green’s strategy of OpenPath is spot on. Build your own direct supply path. And, honestly, I think SSPs are priming to eat Jeff’s lunch. But it’s not like The Trade Desk needs to build a curation product, per se. It’s just that it’s not getting access to a lot of data that it could have.
At the same time, I am surprised at the language Jeff uses to push IDs. He said [during TTD’s Q4 2023 earnings] publishers have a dirty little secret that people aren’t authenticated. Well, you’ve got a dirty little secret that you can only buy on IDs. And The Trade Desk hasn’t done any work to bid on non-ID-based signals, especially publisher cohorts and audiences. So the SSPs have come in and hoovered up that strategy.
If Jeff Green really wants to make this work, he needs to start listening to publishers and their signals. But he’s not. Instead, he’s retrenching in CTV and retail media, and he’s abandoning the open internet.
Given the momentum behind deal curation and persistent media quality concerns in open auctions, is the open internet done as a concept?
No, because the open web performs. But if no one changes how [DSPs only bid on IDs, then] the open web is done.
This interview has been edited and condensed.
For more articles featuring Joe Root, click here.
