Picture one ad slot on one news site, loading for one reader, right now. You might assume there is one way to buy it. There are usually a dozen. The publisher works with several supply-side platforms directly. Each of those SSPs may also be reachable through other SSPs that resold the connection. The same slot, the same reader, the same hundred milliseconds, arrives at a buyer's demand-side platform as ten or fifteen separate bid opportunities, each routed differently, each priced differently, each carrying a different set of fees skimmed along the way.
A buyer who bids on all of them is paying to compete against itself. Supply path optimization is the discipline of refusing to do that. It is the buyer deciding, on purpose, which routes to an impression are worth taking and which to ignore.
Origin: how one impression got a dozen front doors
For the first stretch of programmatic, the route to an impression was reasonably tidy. A publisher ran a waterfall: it ranked its demand sources by their historical average price and called them one at a time, top down, until one accepted the impression. Slow and imperfect, but at least each impression took a single, traceable path.
Header bidding ended that, and it ended it for good reasons. Instead of asking demand sources one after another, a publisher started asking all of them at once, in a simultaneous in-page auction. That fixed the waterfall's worst flaw, where a lower-ranked source willing to pay a premium for one specific user never got the chance. But it had a side effect nobody designed. Once every SSP could see every impression at the same moment, a buyer's DSP started receiving, in the words of AdExchanger's explainer, dozens of bids for the exact same impression. The same inventory, now available through many doors at once.
Reselling widened the problem. An SSP that cannot connect to a publisher directly can buy that publisher's inventory from another SSP that can, then offer it onward. The impression reaches the auction two or three hops deep, and every hop takes a cut. So the buyer is not choosing between different inventory. It is choosing between different routes to identical inventory, and the routes are not equal. One is a clean direct line. Another passes through three intermediaries, each adding a fee and a layer of fog. Same ad, very different price, very different ability to know what you bought.
SPO emerged around 2017 as the direct answer to this, a couple of years after header bidding went mainstream and made the mess. The core idea is almost embarrassingly simple, and Will Doherty of The Trade Desk has argued the industry overcomplicates it: understand who is touching your inventory, and remove as many of those touches as you can.
Present: the transparency tools that make SPO possible
You cannot shorten a path you cannot see. SPO only became practical because the industry shipped a set of plumbing standards that let a buyer trace a bid request from the publisher all the way down. Three pieces matter, and they answer three different questions.
The first is ads.txt, short for Authorized Digital Sellers. It launched in June 2017 through the IAB Tech Lab, built to kill domain spoofing, where a fraudster sells fake inventory dressed up as a premium site. A publisher posts a plain text file at its domain, reachable by adding /ads.txt to the address. The file lists every company authorized to sell that publisher's inventory, each line carrying a seller domain, a seller ID, and whether the relationship is DIRECT or RESELLER. In one sentence: ads.txt is the publisher declaring who is allowed to sell its ads. A streaming-TV variant, app-ads.txt, does the same job for environments without a normal web page.
The second is sellers.json. Ads.txt names the companies a publisher trusts, but it does not tell you who those companies actually are, or who sits behind a reseller line. Sellers.json fills that gap. Released by the IAB Tech Lab in 2019, it is a public file hosted by each SSP and exchange that lists every seller it works with, including each seller's name, domain, and role, whether publisher or intermediary. In one sentence: sellers.json lets a buyer look up who every seller and middleman in a chain really is.
The third is the SupplyChain Object, often written schain. Ads.txt and sellers.json are reference files you check; the SupplyChain Object travels with the bid request itself. It records the actual hop-by-hop route the request took. It is built from a set of nodes, one per company that handled the request, and each node carries an advertising system identifier (asi) and a seller ID (sid). The nodes are chained in the same order as the money flows. In one sentence: the SupplyChain Object is the receipt that shows every hop a single bid request made on its way to you.
Together they let a buyer do something that was impossible a decade ago. Take a live bid request, read its SupplyChain Object to see the full chain, then cross-check every node against the relevant sellers.json and ads.txt files to confirm each link is who it claims to be and was authorized to be there. Index Exchange describes the three as a coordinated system, and that is the point: ads.txt says who may sell, sellers.json says who they are, schain says who actually did. Adoption is now near-universal. Jounce Media's April 2026 benchmarking found 97.6 percent of bid requests linking to fully transparent sellers.json entries across the exchanges it monitors.
How a buyer actually runs SPO
SPO is not a toggle. It is an audit followed by a habit. The practical sequence, drawn from how practitioners describe it, runs roughly like this.
Start with log-level data. This is the granular, impression-by-impression record of what a buyer's DSP did: every bid, every path, every fee, every win and loss. Without it, SPO is guesswork, which is why a recurring instruction in SPO guidance is to work only with partners that pass 100 percent of it. With it, a buyer can pull a report and see, for a given publisher, every distinct path its impressions arrived through.
Then find the duplicates. A single publisher will show up through several SSPs, sometimes a dozen. For each route, the buyer compares the things that actually differ: the take rate, the win rate, the share of invalid traffic, the auction behavior, whether the path is direct or resold. The questions are blunt. Which route to this publisher is cheapest? Which is cleanest? Which ones are just the same inventory wearing a different badge?
Then cut and consolidate. Keep the direct and efficient paths. Drop the rebroadcasters and the high-fee, low-transparency routes. Express the decision through allow and block lists, and through deals negotiated straight with the SSPs the buyer wants to favor, often for a lower fee or preferred treatment in return for the concentrated spend. The Lotame framework lays out the same shape: assess paths from DSP reporting, consolidate onto top performers with allow and block lists, then monitor continuously. That last word matters. New paths appear constantly, so SPO done once decays. It has to be a standing process, not a spring cleaning.
The scale of the cut can be dramatic. A survey of buyers cited by Performance Marketing World found advertisers working with an average of six SSPs after SPO, down from twenty-seven. The Trade Desk took the logic to its end point with OpenPath, a direct integration that connects its DSP straight to publisher inventory with no SSP in the middle at all. A year into OpenPath, The Trade Desk described the result as a clean, unadulterated supply chain.
The benefits, when SPO is done well, are concrete. More of the budget reaches publishers because fewer middlemen skim it, which connects directly to the ad-tech tax, the large share of every programmatic dollar that never reaches a quality viewable impression. Fraud exposure drops, because resold and unauthorized paths are exactly where spoofed and invalid inventory hides. And the data gets cleaner, because a shorter chain is a chain you can actually read.
Future and impact: the reselling era is closing
There are honest limits, and they are not small. Consolidating onto six SSPs instead of twenty-seven means fewer bidders competing for some inventory, which can reduce auction pressure and trim a buyer's reach on harder-to-find supply. Over-aggressive blocking can strip out brand-building inventory along with the junk. And some authorized resellers genuinely do offer access or audience data a direct path does not, so cutting purely by hop count can cost something real. SPO is a trade, not a free win.
It is also no longer a standalone discipline. SPO is blending into programmatic curation, where a buyer or a trusted partner pre-assembles a marketplace of vetted publishers and strips out the resellers and low-quality inventory before the auction runs. Curation is, in effect, SPO packaged as a product. The line between choosing your paths and buying a pre-cleaned set of them is fading.
The sharpest signal comes from the supply data itself. Jounce Media's benchmarking has been tracking rebroadcasting, its term for resold supply chains, and the numbers describe an era ending. In the May 2026 report, rebroadcasting accounted for 27 percent of all bid requests but only 18 percent of total spend. Per request, resold supply monetizes at roughly a 50 percent deficit against maximally direct paths. DSPs are not waiting around: The Trade Desk has put a near-comprehensive platform-wide block on rebroadcasting, and other DSPs are following. When buying machines systematically pay direct paths more and resold paths less, reselling stops being a viable business.
The companies that lived on reselling are now pivoting, mostly toward curation and toward managed wrappers, where payment routes directly from an exchange to both the publisher and the wrapper provider. That fixes the optics without fully fixing the duplication. As Jounce notes, a publisher running six managed wrappers can look 100 percent direct on paper while still generating enormous duplicate volume, which is why the IAB Tech Lab has a schain extension and a deals API in progress to expose wrappers and curators the way ads.txt once exposed resellers. The transparency standards are chasing the next hiding place.
The throughline holds, though. Chris Kane of Jounce Media has framed the new economics plainly: in a survey of buyers rating the major programmatic sources, the publishers and paths with the highest trust scores captured more than half of all spend despite carrying a smaller share of bid requests. Direct integration is the strongest predictor of where the money goes. Buyers are also shifting from sprawling exclusion lists toward inclusion lists built around a known set of trusted publishers. The logic, as Kane has put it to AdExchanger, is one of scale: the open internet looks like 1.5 million properties, which makes picking the good ones hard, but on a seller-by-seller basis there are at most a thousand sellers that matter, and more like a hundred.
That is the real lesson of supply path optimization. The dozen doors to an impression were never equal. SPO is just the buyer finally reading the receipts, and the market is now pricing the lesson in for everyone.
Council summary
This post argues that supply path optimization is not a clever tactic but a correction: header bidding and reselling turned one impression into a dozen unequal routes, and SPO is the buyer's discipline of buying only the good ones. It earns its length by teaching the three standards that make the work possible, ads.txt for who may sell, sellers.json for who they are, and the SupplyChain Object for who actually did, then walking through the audit, the deduplication, and the standing habit that real SPO requires. The reader's takeaway is concrete: trace your paths from log-level data, consolidate hard, and treat it as continuous maintenance, because the duplication keeps regrowing in new forms. The honest section on the cost of over-consolidation keeps it from reading as a sales pitch, and the closing data, resold supply now monetizing at a deficit and DSPs blocking it outright, lands the real point. Reselling is not being argued out of the market; it is being priced out.
Comments