Everyone is de-layering. Almost nobody is deciding.

Every buying team I talk to is shortening something this year. Fewer SSPs on the plan. Direct lines to the publishers that matter. Curation brought in-house. The word for all of it is de-layering, and the promise is control.

The promise is fair as far as it goes. The chain got long, the well-worn studies say roughly half of a programmatic euro reaches the publisher, and nobody could see where the rest went. Cutting layers is a reasonable answer to that.

But watch what happens after the layers are cut. The path is shorter, the fee line is better, the plan looks cleaner. And the campaign performs about the way it did before. Not worse. Not much better either.

That is not a failure of de-layering. It is de-layering doing exactly what it does, and nothing else.

What everyone's saying

Three versions of the same pitch, and you'll hear all of them this quarter.

The efficiency pitch: cut the resellers, cut the duplicate paths, and more of the budget lands in media. True, and measurable.

The visibility pitch: with fewer hands on the request, you finally see who touched it and what they took. Also true, and long overdue.

The control pitch: shorten the path and you take the buy back from the platforms. Decide for yourself instead of accepting what the chain hands you.

The first two are about cost and sight. The third is about judgment. And that is where the word starts covering more ground than the thing it describes.

What de-layering actually does

Strip the pitch and de-layering is plumbing work. You remove intermediaries between a bid request and a publisher. Every hop you remove takes its fee with it, and every duplicate path you close stops you bidding against yourself for the same impression.

That is worth doing. It is also finite. Once the path is short, it is short. There is no version of the same campaign that gets better because the request travelled through two hops instead of four.

Here is the part the control pitch skips. Somewhere on that shorter path, a decision is still being made about which impressions deserve your bid. Which sites are in. Which contexts count. What bid CPM the deal deserves. Who made that decision, when, and on what evidence, is a completely separate question from how many companies sat between you and the publisher.

Cut every layer you like. The decision is still there, still made by someone, still made at some moment. De-layering changes the route. It does not change the choice.

Why it matters when you're spending the budget

Here is what "keeps deciding" looks like in most plans today, and it is only honest to say it: the DSP does it. Once the deals are live, its algorithm watches the flow of requests carrying each deal ID, learns which ones clear and perform, and throttles the rest. That is a decision made continuously, after launch, by a machine doing its job well.

It is also a reaction, not a choice. The DSP decides how hard to bid on what it is fed. It never decides what it should have been fed.

And it is fed less than most buyers think. Before a request reaches the DSP, the SSP has already shaped its traffic, sending each bidder more of what it tends to buy and less of the rest, because nobody can afford to listen to everything. So there are two throttles on every plan, one before the DSP and one inside it, and neither of them is a selection. A partner of Bedrock, the containerised DSP incubated by FirstPartyCapital, put it plainly when they moved their bidder inside Index Exchange's cloud this spring: "If you can only see a fraction of the market, your models are inherently constrained." That is the honest state of the buy. The smartest algorithm in the chain works on a sample it did not choose.

Now look at the levers a trader actually pulls after launch. Bid CPM, per line and per deal. Budget between lines. Now and then a deal switched off. What almost never moves is the composition of what carries the deal ID: which sites, which contexts, on which SSP. That was set at launch, usually by whoever packaged the deal, and the DSP spends the next six weeks throttling around it.

Two things follow. A deal the DSP keeps throttling is a deal that was not built for this brief. Raising the bid CPM buys a little more of it; it does not make it the right deal. And consolidation genuinely helps here, which is the honest case for de-layering: fewer duplicate paths means cleaner bid density and better win rates. That is exactly why it gets mistaken for a better decision. The win rate went up. The selection did not change.

A test for Monday. Pull the deals live on your biggest campaign and note two things for each: has its composition changed since launch, and what has the DSP done with it (bid rate, win rate, share of budget). Most teams find that the only things that moved are bid CPM and budget, that the DSP has quietly starved a share of the deals, and that nobody re-cut a single one. Short path, live throttle, frozen selection. That is what de-layering leaves you with.

The counter-example exists, and it is measurable. On one multi-market video campaign we documented, 28 deals built for the objective across 7 destinations delivered a 37.8% lower cost per view than the rest of the same campaign. Same DSP, same budget, same path. The whole difference sat in what carried the deal ID.

If you run the business rather than the campaigns, the same fact reads like this: the fee saving from a shorter path shows up once, on a rate card. The value of a better selection shows up every time the client renews. You are being shown the first and asked to trust the second.

What nobody pushing de-layering will admit

Two things, and both are uncomfortable.

The first is that the decision has moved further from the buyer, not closer, while the chain got shorter. Most programmatic now runs through deals assembled in advance: Jounce counted 3.5 deal IDs riding on an average bid request, and under one request in five reaching the open auction as it was designed. The path to the publisher is short. The selection inside the deal was made before your campaign started, by whoever packaged it, and the only thing revisiting it since is a throttle. Shorter path, older decision.

The clearest case is not in Europe. When The Trade Desk rolled out SP500+, it put an include list of publishers at the system level and switched it on by default. Ari Paparo's reading is the right one: the DSP became the club bouncer, deciding which supply it lets itself be shown. That is de-layering done for you, by the platform, on a decision you did not make and most teams have not looked at. If you buy through that seat, the first thing to check is not your fee line. It is whether the list is on, and who put what on it.

The second is what happens next. Agents are coming to the buy, and everyone selling one will tell you they close the loop. They will transact faster than any trader, across any path. But an agent buys from what it is given to choose from. Point it at a pool selected once and never re-cut, and it will do what the DSP already does, faster: optimise brilliantly inside a decision that was already wrong. Speed does not fix selection. It amplifies it.

To be fair, a few teams are doing the harder thing: re-cutting what carries the deal ID while the campaign runs, letting what delivery teaches change the selection the next morning, on their own seats or on a partner's. It exists. It is rare. And it is what "control" should have meant all along.

So here is the question to carry into the next conversation, with a partner or with your own team. Not "how many hops are on this path." That one has an easy answer and it is aimed at the smaller prize.

Ask: since we launched, what has changed in what carries our deal IDs, and who changed it?

If the honest answer is "the DSP throttled some, and nobody touched the rest," the path is short and the decision is frozen. You have de-layered. You have not started deciding.

The real question nobody's asking

De-layering was supposed to solve a trust problem. The chain was long and opaque, so the industry set out to make it short and visible. It worked, as far as it went. Fewer hops, better fee lines, a plan you can explain.

But look at what was fixed: the route. And look at what the outcome of a campaign actually depends on: the choice of what to buy, and whether anyone keeps making that choice once the money is moving. Chris Kane at Jounce said it years ago, looking back at the first wave of SPO: marketers care only indirectly about take rates; what they care about is return. The teams that took that seriously, and worked the selection instead of the fee line, are the ones he found improving ROI by 10% or more.

Which is the pattern under every word in this series: in adtech, the plumbing is almost never the problem. The decision always is. Nobody struggles to shorten a path anymore; the tools do it for you. What almost nobody does is decide, for this brief, which inventory deserves the bid, and keep deciding while the campaign runs.

That is the control worth taking back. Not a position in the chain. A decision, made on purpose, made for this campaign, and made again tomorrow.

---

Also in this series:

• Episode 01: what "agentic" actually means
• Episode 02: MCP, the plumbing everyone name-drops
• Episode 03: AdCP, briefing inventory in plain English
• Episode 04: curation, a decision made before your campaign started

Maxime Khalfallaoui, Supply Finder. I came up on the SSP side. Now I help teams that sell media outcomes deliver what they sold.