What is bid duplication, and how do buyers handle it?
Part of Supply path transparency
The short answer
Bid duplication is when one ad opportunity reaches a buyer as several bid requests, through different exchanges, resellers or integrations. It inflates the supply a buyer appears to see, spends bidding capacity on copies and can make a buyer compete with itself. Buyers handle it by grouping copies, choosing a preferred path and asking sellers to stop pure clones.
Key takeaways
- One page view can become several bid requests because selling platforms can call further selling platforms before bids return.
- IAB Tech Lab's 2026 draft separates variation, copies that differ in path or content, from replication, copies that are exact clones, and asks sellers not to replicate.
- Duplicates make supply look larger than it is and, in second price rules, can let a buyer's own copy set its price.
- Buyers deduplicate with shared identifiers, the SupplyChain object and a preferred path per publisher.
- The exchange's own request ID cannot match copies, because OpenRTB lets it differ for each recipient.
What is bid duplication in programmatic?
Definition
- Bid duplication
- The arrival of one underlying ad opportunity at a buyer as more than one bid request, whether through different selling paths or as repeated requests from the same seller.
Trade writing uses several names for the same thing: duplicate bid requests, request duplication, auction duplication. IAB Tech Lab's Programmatic Best Practices (opens in a new tab), released for public comment on September 16, 2026, gives the clearest primary definitions. It says bid requests "should be sent to each partner only once per underlying ad opportunity and per integration mechanism or inventory partner domain," and splits duplication into two categories.
| Pattern | The draft's words | What a buyer sees | The draft's guidance |
|---|---|---|---|
| Variation | Requests for the same opportunity that "traverse different supply chains or paths, contain materially different information in each bid request, or both" | The same slot from two exchanges, or with different deals or data attached | Accepted where differences are material, such as privacy or transparency requirements |
| Replication | Requests that "are exact clones or the variance is not material" | Identical requests arriving more than once | "Do not engage in replication." |
| Fan out | "exploding arrays into individual bid requests or giving every deal_id a separate bid request" | One request per deal or per format for a single slot | Avoid it where integration partners support arrays |
The draft also concedes some duplication is unavoidable where the current OpenRTB spec cannot express value, and asks that it be "constrained, incremental, and made observable to downstream systems." It was still labeled public comment when checked on October 22, 2026, so treat it as the direction of the standard rather than a final rule.
Why does one impression arrive through several SSPs?
Because the selling side is built in layers. IAB Tech Lab's Programmatic Auction Definitions (opens in a new tab) describe a request initiator, such as a header bidding wrapper, that calls selling platforms, and say "each selling platform may call (n) additional selling platforms in a single ad request initiation." The process "could happen recursively" at each hop. They add that "Intermediary auctioneers may forward more than one accepted bid to upstream partners."
Several ordinary setups produce copies. A publisher may run a client side wrapper and a server side one that both call the same exchange. A reseller authorized in the publisher's ads.txt file may offer the same slot that the publisher's direct account offers. A seller may send one request per deal instead of listing the deals together. None of these needs bad intent, which is why the buyer's task is sorting copies, not blocking sellers. The header bidding guide covers the wrapper side in detail.
How does duplication affect a buyer's bids?
The first effect is on what the buyer thinks it sees. A DSP counting bid requests per domain will overstate available supply when each slot arrives three times. Forecasts, reach estimates and win rate reports built on request counts inherit that error. A site can look large and hard to win when it is small and arriving through several doors.
The second effect is on capacity. Each copy has to be read, evaluated against targeting and answered or ignored inside the auction's time limit. Tech Lab's draft states the aim plainly: by reducing "redundant auctions for the same ad opportunity, volume and processing costs are reduced throughout the supply chain," and platforms can focus on advertiser goals. A bidder that throttles incoming traffic to cope may drop the copy from the path it prefers and keep a worse one.
The third effect is on pacing and frequency. A campaign that answers each copy may place several bids for one opportunity. Only one ad can show, but budget reservations and frequency logic may treat the copies as separate chances, depending on how the platform is built.
Does duplication raise prices?
It can, depending on where the copies meet and which pricing rule applies. The MRC's January 2026 Digital Advertising Auction Transparency Standards (opens in a new tab) define a first price auction as one where the "auction winner pays their bid" and a second price auction as one where the winner "pays the second-highest bid (or a penny more)." The table works through what follows when a buyer's own copies compete. These are consequences of the definitions, not measured outcomes.
| Where the copies meet | Pricing rule | Possible effect on the buyer |
|---|---|---|
| In one final auction | Second price | The buyer's other copy can be the second highest bid, so the buyer pays close to its own top bid |
| In one final auction | First price | The buyer pays the higher of its own bids; the duplicate does not set the price but can raise the winning bid |
| In separate auctions on different paths | Either rule | Each path clears on its own; the cost shows up as bidding capacity and fees on the path that wins |
| Inside a deal | Deal terms | OpenRTB lets a deal set at to 3, where the bidfloor value is the agreed deal price, so each copy carries the same agreed price |
The MRC standards also tie the fix to bidding behavior. They make the transaction ID a minimum requirement "contingent on DSP adoption of Multi-Bid," and describe multi-bid as a DSP sending several bids for the same impression in a single auction instead of one best bid. Tech Lab's draft makes the same link from the seller side, asking demand to support multi-bid responses "especially when TID is present in the bid request." The logic is that sellers fan out less when buyers can answer one request with several options.
How do buyers deduplicate bid requests?
Deduplication needs a key that is the same on each copy. The obvious candidate fails: the OpenRTB 2.6 (opens in a new tab) BidRequest id is "assigned by the exchange" and "The exchange may use different values for different recipients." The Source object's tid is defined as a "Transaction ID that must be common across all participants in this bid request," when sellers send it. The transaction ID guide covers how it is set and where it goes missing.
Five steps to group copies and pick a path
- Group by the shared ID where present. Treat requests that carry the same transaction ID as one opportunity, whichever exchange sent them.
- Fall back to a composite key. Where no shared ID arrives, combine domain, placement identifier, slot size and a short time window, and accept that the match is approximate.
- Read the path for each copy. The SupplyChain object (opens in a new tab) lists each node, and each node's rid is "The OpenRTB RequestId of the request as issued by this seller."
- Choose a preferred path per publisher. Favor the copy whose chain is complete and starts with the publisher's own direct account, then bid on that copy and pass on the rest.
- Watch requests per impression won. Track it by domain each month; a sudden rise points to a new path sending copies.
Platforms also act on their own. Display & Video 360 announced (opens in a new tab) in its September 23, 2024 edition that "Starting October 23, 2024, Display & Video 360 will stop bidding on multiple waterfall calls to Google Ad Manager," extending the change to the AdMob Network and mediation platforms for mobile app inventory.
What does duplicate path analysis look like on a real plan?
Fictional example
A roller skating rink chain sorts its copies in Alabama and Mississippi
A trading desk runs display ads for an unnamed chain of roller skating rinks promoting birthday parties and family nights in Birmingham, Huntsville and Jackson. The plan lists 45 English language sites: local news, parenting, family events and school sports.
Request logs show one parenting site sending far more requests than its size suggests. Grouping by transaction ID shows each slot arriving four times: twice from the same exchange, once through a client side wrapper and once through a server side one, and twice more through two resellers. The two requests from the same exchange are near identical.
The desk keeps the copy from the publisher's direct account, whose chain is complete, and stops bidding on the reseller copies for that site. It sends the exchange a note asking whether the two integrations are meant to send identical requests, quoting the replication guidance in Tech Lab's draft. For three school sports sites that send no transaction ID, it uses domain, slot and time as an approximate key and flags them for review.
How does PlacementGraph help before path analysis starts?
What should buyers ask sellers about duplicate requests?
Questions for an exchange or reseller
- Do you send each buyer one request per opportunity for each integration, or several?
- Do you send separate requests for each deal ID or format on one slot, instead of listing them in one request?
- Do you pass the transaction ID you receive without changing it?
- Do you add your own node to the SupplyChain object and keep the complete flag accurate?
- Do you accept multi-impression and multi-format requests from publishers, so they do not need to split them?
- Do you declare ad refresh in the request, so a refreshed slot is not mistaken for a copy?
Frequently asked questions
Is bid duplication a type of ad fraud?
Not by itself. Copies from authorized paths are a byproduct of how selling platforms connect. Fraud involves misrepresenting inventory or traffic, which the ad fraud field guide covers. Duplication becomes a concern when copies come from unauthorized sellers or when pure clones add cost without adding anything a buyer can use.
Can publishers reduce duplication on their side?
Yes. Tech Lab's draft asks sellers to send each partner one request per opportunity per integration mechanism or inventory partner domain, to include a SupplyChain object in each request, and to avoid exploding arrays into separate requests. A publisher can also review whether a client side and a server side setup both call the same exchange.
Do private marketplace deals avoid duplicate requests?
Not automatically. The draft names "giving every deal_id a separate bid request" as a form of fan out. A slot eligible for three deals can arrive as three requests unless the seller lists the deals together. Deal buyers can ask the exchange how it packages deals before the campaign starts.
Why would the same exchange send two requests for one slot?
The publisher may have integrated that exchange more than once, for example in two wrappers, or the exchange may split formats or deals into separate requests. The draft allows one request per integration mechanism, so two integrations can be legitimate. Identical copies with no material difference are what it calls replication.
Sources
- Programmatic Best Practices version 1.0 (public comment, September 16, 2026) (opens in a new tab)IAB Tech Lab
- Programmatic Auction Definitions (opens in a new tab)IAB Tech Lab
- OpenRTB 2.6 specification (opens in a new tab)IAB Tech Lab
- SupplyChain object specification (opens in a new tab)IAB Tech Lab
- Digital Advertising Auction Transparency Standards (January 2026) (opens in a new tab)Media Rating Council
- Display & Video 360 announcements, September 23, 2024 edition (opens in a new tab)Google Display & Video 360 Help
Researched and written by PlacementGraph Research. Read how we research, source and update these pages. Corrections: diana@placementgraph.com.