What is curation in programmatic advertising?

Curation is when someone pre-selects publisher inventory, often adds audience or contextual data, and bundles it into a deal you buy through your DSP. Instead of bidding across the open exchange, you buy from a pool someone else assembled. The curator is paid when their deal wins an impression.

Is curated inventory better than the open exchange?

Usually yes on quality, because fraud, made-for-advertising sites and low-attention placements get screened out before the request reaches you. But the result depends on who did the selecting and what they were optimising for. Curation is a mechanism, not a guarantee.

Does a curated pool update while a campaign runs?

Usually not. Most pools are assembled at a point in time against the criteria that applied then, and stay as they are while the campaign runs. A handful of players are starting to close that loop with algorithmic curation, which is exactly why the question is worth asking your partner directly. The answer tells you a lot.

Episode 04 · 7 min read

Curation: you're buying a decision made before your campaign started

Key takeaways

Curation is a tool, not a service someone sells you. It bundles inventory into a buyable deal, and every major exchange ships one.
Two curators using the same platform hand you different pools. What separates them is access and judgment, rarely the tooling: everyone's tooling runs the same filters.
Almost every pool shares the same limit: it was assembled once, in advance, and nothing that happens during your flight feeds back into it.
The selection drives far more of your outcome than the fees do, and it is the part of your work nobody downstream can see.
For the person running the agency: the quality of the buy is locked before the first euro serves, and nobody in the chain can show the client the work behind it.
Ask whether a pool adjusts while your campaign runs, or whether it is the list built the day it was sold to you. Usually it is the list.

Look at the deals running in your plan right now. Most of them were assembled weeks ago, by someone else, against criteria you never saw. And they haven't changed since.

That isn't sloppiness on anyone's part. It's how the market works now. Somewhere between the open auction you learned on and the plan you signed off last week, buying became a matter of picking from pools other people packaged in advance.

Nobody put that to a vote. Both sides had reasons to want it: buyers pushing for verified, brand-safe environments, and platforms building the packaging to deliver them. The open auction went from default to exception without most buyers marking the moment.

So "should I engage with curation" is the wrong question. You're in it. The useful question is what you're actually buying when you buy a pool.

What everyone's saying

Three pitches, and you'll hear all of them this month.

The quality pitch: curation is how you escape the open exchange. Fraud, junk sites and low-attention placements screened out before the request ever reaches you.

The data pitch: curation is how audience signal gets attached to inventory at the source, so you buy the right people in the right place instead of hoping your targeting sorts it out later.

The control pitch: curation is how you shorten the chain between you and the publisher, so more of the budget lands in actual media.

All three describe something real. All three are also things a vendor can say while doing something quite different, and the word covers enough ground that you'd never know which.

What it actually is

Strip the packaging and curation is a tool. Someone uses it to bundle a pool of inventory, wrap it in a deal, and make it buyable through a DSP. The data layers, the viewability floors, the attention thresholds: those describe how they used it.

And the tool is everywhere. Every major exchange ships one now. It stopped being a specialist capability years ago and became a standard feature of the stack.

Which includes you. Nothing structural stops a buying team from holding the pen on its own pools, and the larger ones have been doing exactly that for years, assembling supply straight from publishers and cutting the hops in between.

So the interesting question was never who's allowed to curate. The tool is commodity, and the filtering it performs is converging across the market. The interesting question is when the selection happened, and what has updated it since.

Why it matters when you're spending the budget

Here's the uncomfortable part of the job. The selection drives far more of your outcome than the fees do. Which pool you buy, and what's actually inside it, moves performance more than any rate negotiation you'll run this quarter.

You know this. You've watched two deals with near-identical CPMs deliver completely different results, and you could explain exactly why if anyone asked.

Nobody asks. What the account team sees is a line in a plan and a CPM. What the client sees is a deal name and a promise of premium. The judgment you applied, the sources you excluded, the reason you paid three euros where someone else paid one fifty: none of it survives the trip from your DSP to the review meeting.

So the part of your work that matters most is the part you have the least ability to show. And when a client pushes back on price, you're defending a decision with a vocabulary that was never built to explain it.

Curation was designed by sellers for technical buyers. It was never designed to be explained to the person paying for the campaign. That's not a detail, it's why good selection work stays invisible.

And if you run the agency rather than the campaigns, the same fact reads differently. The quality of every programmatic buy your teams make is locked in before the first euro serves, by a selection nobody on your side can audit or show the client. When the review meeting turns to media quality, your margin is being defended with a deal name and a CPM. The work that justifies both is sitting in a place the client will never see.

What nobody selling you curation will admit

Two curators working from the same platform, applying the same stated thresholds, won't hand you the same pool. What separates them is rarely technical. It's their access to publishers, what they know about their inventory, how they read quality. Picking the right partner genuinely matters.

But almost every pool shares the same limit, whoever built it: it was assembled once.

Someone put it together on a Tuesday, against the criteria that mattered that Tuesday. Then your campaign runs for six weeks. That pool doesn't know which placements underdelivered last week. It doesn't tighten when a source starts slipping. It's a list, frozen at the moment of packaging, sitting inside a market that moves by the hour.

That isn't a failure of curation. It's what curation is: selection, done in advance. But it means the quality of your buy was set before a single impression served, and nothing you learn during the flight feeds back into it.

To be fair, a few players are starting to close that loop, rebuilding pools algorithmically as the campaign teaches them things. It exists, it's rare, and most of what's sold under the word today doesn't do it. Which is exactly why you should ask rather than assume.

So here's the question to carry into the next conversation, and it isn't the one you'd expect. Not "what does this cost." That one has an easy answer and it's aimed at the smaller prize anyway.

Ask: does this pool adjust while my campaign runs, or is it the list you built the day you sold it to me?

Most of the time, it's the list. Worth knowing before you commit six weeks of budget to it.

The real question nobody's asking

Curation was supposed to solve a visibility problem. The open exchange was opaque, so the industry built tooling that could sort the good from the bad before the request went out.

It worked. The tooling spread everywhere, the pools got cleaner, and buying got better for it. But look at what the fix actually was: a better decision, made earlier, and then frozen.

Which is the pattern under every acronym in this series: in adtech, the technology is almost never the problem. The operational layer always is. Nobody struggles to assemble a good pool. What almost nobody does is keep deciding after the deal goes live, as the campaign teaches you things the packaging never knew.

That's the gap. Not fees, not filtering, not who gets to curate. The gap sits between a selection made once and a market that moves continuously. Everything you learn during a flight currently dies in a report nobody reads, when it should be shaping what you buy tomorrow morning.

The word will keep drifting. It's too useful to too many people in its current vagueness. But underneath it sits something simple: someone assembled this, on purpose, at a moment. Know when that moment was, and whether anything has updated it since.

Also in this series:

Maxime Khalfallaoui, Supply Finder. I came up on the SSP side. Now I help agencies find their way through it.

Questions people actually ask

What is curation in programmatic advertising?: Curation is when someone pre-selects publisher inventory, often adds audience or contextual data, and bundles it into a deal you buy through your DSP. Instead of bidding across the open exchange, you buy from a pool someone else assembled. The curator is paid when their deal wins an impression.
Is curated inventory better than the open exchange?: Usually yes on quality, because fraud, made-for-advertising sites and low-attention placements get screened out before the request reaches you. But the result depends on who did the selecting and what they were optimising for. Curation is a mechanism, not a guarantee.
Does a curated pool update while a campaign runs?: Usually not. Most pools are assembled at a point in time against the criteria that applied then, and stay as they are while the campaign runs. A handful of players are starting to close that loop with algorithmic curation, which is exactly why the question is worth asking your partner directly. The answer tells you a lot.