What is a trading desk, and what does it check during a campaign?
Part of Supply path transparency
The short answer
A trading desk is the team that operates programmatic buying for an agency, a brand or several clients. It sets up campaigns and deals in buying platforms, sets bids and pacing, watches delivery, and checks that impressions come through authorized sellers. It turns a planner's approved site list into live buying and reports back.
Key takeaways
- A trading desk is a team, not a platform: it runs the buying platform seats where bids are made on a client's behalf.
- Daily work is a loop of delivery, deal health, filtered traffic, pacing and discrepancies, read from the buying platform's own diagnostics.
- A one page evidence sheet per domain keeps the seller facts, deal IDs and decisions that a client or auditor will ask for.
- Planners own which sites belong on a plan and why; the desk owns how they are bought, and both share one change log.
What is a trading desk in advertising?
Definition
- Trading desk
- A team of programmatic traders that buys digital media through buying platforms on behalf of advertisers. It configures campaigns, deals, bids and targeting, monitors delivery and reconciles what was bought with what was planned.
OpenRTB does not use the term, but it describes the job. The OpenRTB 2.6 specification (opens in a new tab) defines a bidder as an entity that competes in real time auctions to acquire impressions, and a seat as an advertising entity, such as an advertiser or agency, that uses bidders to act on its behalf and usually owns the budget. A trading desk is the team that runs those seats; the MRC's auction standards call this role an advertiser or their representative bidding into an auction.
Desks are organized in one of three ways: a team inside an agency group, an independent company that buys for agencies or brands, or a brand's own in house team. Each works; what changes is who holds the seat, who sees the log data and who answers for supply quality.
| Model | Who runs the seat | Strengths | Points to settle in the contract |
|---|---|---|---|
| Agency trading desk | A team inside the agency group | Close to planning and to the client relationship; shared tools across clients | Who owns the seat and the log data if the client moves agencies |
| Independent trading desk | A separate specialist company | Focused trading skills; can serve agencies of any size | How fees are shown, and how reports reach the planner and the client |
| In house team | The advertiser's own staff | Direct access to data and decisions; no extra party in the chain | Staffing, training and coverage when people are away |
What does a trading desk check during a campaign?
The in flight routine is diagnostic: is each line item and deal finding inventory, bidding, winning and spending as planned, and if not, where is it losing? Buying platforms expose this directly. Google's Display & Video 360 Troubleshooter (opens in a new tab), for example, shows total available requests, bid responses, impressions won and win rate for a line item, with requests, impressions won and win rate available within one hour, and breaks lost availability into filtered impressions, targeting, creatives, frequency, budget, pacing and auction losses.
A routine check for one campaign
- Read delivery against pacing. Compare spend and impressions with the flight plan for each line item.
- Open each deal's volume. In Display & Video 360, the Delivery column under My Inventory (opens in a new tab) shows how much inventory each deal gets and how often the buyer wins it.
- Read the filtered traffic. Invalid traffic, unauthorized seller under ads.txt, or a request sent before the deal was created each need a different fix.
- Read the auction losses. A bid below the deal floor or the exchange floor is a price problem; a loss to another campaign in the same account is an internal conflict.
- Check creatives. Rejections for size, format, secure serving or a restricted category stop delivery before price matters.
- Note every change. Log each bid, floor, pacing or targeting change with the date and reason.
Discrepancies are part of the read. The OpenRTB implementation notes (opens in a new tab) explain that winning an auction does not guarantee delivery, that most exchanges bill on some form of delivery or rendering, and that bidders are strongly advised to track the time between the auction and the billing notice. IAB Tech Lab's programmatic auction definitions (opens in a new tab) note that most auction participants reconcile numbers between systems monthly.
How does a trading desk review supply paths?
Planners check who may sell a site before it reaches the plan. A desk checks who actually sold each impression once the campaign runs, comparing ads.txt and sellers.json with logged bid requests and the SupplyChain object, or schain, that rode along with each one.
The MRC's January 2026 Digital Advertising Auction Transparency Standards (opens in a new tab) give a desk a useful list of what to ask its buying platform for. To get the most from supply chain reporting, the standards say DSPs are required to validate seller IDs declared in ads.txt against the SSPs' sellers.json files, to ensure the supply chain ends at the owner of the inventory, and to give buyers transparent reporting on the average number of nodes, or hops, and on mis or undeclared supply chains.
Supply path review of a live campaign, by the desk
- A log sample was pulled for every domain that spent, with seller ID, exchange, schain and deal ID kept.
- Each seller ID was matched to the domain's ads.txt line and to the exchange's sellers.json entry.
- Incomplete supply chains and chains that do not start with the site's owner were listed by exchange.
- The average number of hops per exchange was requested from the buying platform.
- Reseller paths were labeled with the company that resells, not removed by default.
What evidence should a desk keep for each domain?
Questions about a site come back to the same facts: who sold it, through which accounts, under which deal, and what the desk decided. One sheet per domain answers a client or auditor in minutes. The fields come from the ads.txt 1.1 specification (opens in a new tab), the sellers.json specification (opens in a new tab) and the campaign's logs.
| Field | What to record | Where it comes from |
|---|---|---|
| Domain and date checked | The exact domain bought, and the day each file was read | The plan and the review date |
| Authorized accounts | Each exchange domain, account ID and DIRECT or RESELLER label | The site's ads.txt lines |
| Owner and manager | OWNERDOMAIN, and MANAGERDOMAIN if a primary or exclusive partner sells the site | ads.txt 1.1 variables |
| Seller identity | seller_type (PUBLISHER, INTERMEDIARY or BOTH), name and domain, or a note that the seller is confidential | Each exchange's sellers.json |
| Paths seen in the logs | Seller IDs and exchanges that actually delivered, with schain complete or not | Log sample from the buying platform |
| Deals | Deal IDs, auction type and floor for each deal that includes the domain | Deal setup and bid requests |
| Open questions | Anything unmatched, with the request id and who was asked | Desk notes |
| Decision | Keep, limit to certain paths, or raise with the planner, with the reason and the date | Desk and planner, together |
How does a trading desk work with planners?
The cleanest split gives planning the question of what to buy and trading the question of how. Problems start when the line blurs: a desk quietly drops slow sites, or a planner adds sites without saying which deals they belong to. An ownership table agreed at kickoff sets the line.
| Decision | Planner | Trading desk |
|---|---|---|
| Which sites are on the list, and why | Owns it, with a reason per site | Flags delivery or seller problems |
| Suitability and avoid topics | Sets them with the client | Applies them in platform settings |
| Deals, seats and line items | Says which sites need a deal | Requests, accepts and sets them up |
| Bids, floors and pacing | Sets budget and goals | Owns daily bidding and pacing |
| Removing a site in flight | Approves it | Proposes it with evidence |
The hand off itself is a file. The planner sends the approved list, the deal sheet and notes on sites to buy only on certain paths. The desk returns weekly delivery by domain and evidence sheets for anything it wants to change. Guides to what a DSP does and curated deals cover the platform side.
What does a trading desk's week look like on a real campaign?
Fictional example
An independent desk buys for a pest control company in Florida
A fictional Tampa pest control company runs a spring display campaign in Florida, focused on Tampa, Orlando and Miami, in English and Spanish. Its agency hands a list of 35 home, lawn and garden and local news sites to an independent trading desk, with a deal sheet for two private marketplace deals.
On day one, one deal shows healthy available requests but few bids, and many filtered impressions are marked unauthorized seller under ads.txt. The evidence sheet for the largest site explains it: the deal's exchange appears on a RESELLER line under a seller ID that does not match the one in the bid requests. The desk sends request ids to the exchange and tells the planner.
By the end of the week the seller ID is corrected, the deal is spending, and every domain that spent has an evidence sheet. A local news site with an incomplete supply chain stays on the list, marked for another log sample next month.
How does PlacementGraph support a trading desk's site lists?
The desk's evidence starts with the list it receives, and that list can arrive with selling facts attached.
Frequently asked questions
Is a trading desk the same as a DSP?
No. A demand side platform is software that bids in auctions. A trading desk is the team that operates seats in one or more of those platforms, deciding bids, deals, pacing and targeting. In OpenRTB terms, the platform is the bidder and the desk acts for the seat whose budget is being spent.
Can a brand run its own trading desk?
Yes. Some advertisers run seats in buying platforms with their own staff, for all programmatic buying or alongside an agency. The work is the same: setup, daily delivery checks, supply path reviews and reconciliation. What changes is who holds the seat and the log data.
Why would a buying platform filter impressions as an unauthorized seller?
Because the seller in the request is not authorized in the site's ads.txt. Display & Video 360's Troubleshooter counts queries filtered because they come from unauthorized sellers, and a line item can also exclude inventory without an ads.txt file. When a seller ID is missing from the file, the publisher or exchange fixes it.
What does reconciliation mean for a trading desk?
It is the regular matching of delivery and spend between systems, such as the buying platform, the exchange and the ad server, and resolving differences. IAB Tech Lab notes most auction participants reconcile monthly, and OpenRTB advises tracking delays between the auction and the billing notice.
Sources
- Troubleshoot your deals and line items (opens in a new tab)Google Display & Video 360 Help
- Monitor impression volume in your private deals (opens in a new tab)Google Display & Video 360 Help
- MRC Digital Advertising Auction Transparency Standards (opens in a new tab)Media Rating Council, January 2026
- OpenRTB 2.6 specification (opens in a new tab)IAB Tech Lab (GitHub)
- OpenRTB 2.6 implementation notes (opens in a new tab)IAB Tech Lab (GitHub)
- Programmatic Auction Definitions (opens in a new tab)IAB Tech Lab (GitHub)
- ads.txt 1.1 specification (opens in a new tab)IAB Tech Lab (GitHub)
- sellers.json specification (opens in a new tab)IAB Tech Lab (GitHub)
Researched and written by PlacementGraph Research. Read how we research, source and update these pages. Corrections: diana@placementgraph.com.