Bid shading is the deliberate submission of a bid lower than the maximum an advertiser would pay for an impression. It exists because of a change in how digital advertising inventory is priced. In a second-price auction the highest bidder wins but pays only slightly more than the runner-up, so bidding the full private valuation carries no penalty. In a first-price auction the winner pays exactly what was bid. A buyer who values an impression at 10 dollars per thousand impressions and bids 10 dollars, where the next-best offer was 3 dollars, hands over the entire difference. Shading closes that gap by predicting the lowest price that still wins, then bidding marginally above it.

The term comes from auction theory rather than the advertising industry. Economists have documented the behaviour in Federal Communications Commission spectrum sales, oil deposit leases, cattle auctions and Treasury bill tenders. Its formal basis traces to William Vickrey's 1961 paper in the Journal of Finance, which established that truthful bidding is dominant under second-price rules and that bidders in sealed first-price formats rationally bid below what an item is worth to them.

How the calculation works

The mechanism sits inside the demand-side platform, the software an agency or advertiser uses to buy impressions. It runs between the moment a value is assigned to an impression and the moment a bid leaves the server, within a few tens of milliseconds.

The bid request arrives in OpenRTB, the IAB Tech Lab specification governing most automated transactions. Two fields set the constraints. The at attribute declares the auction type, where 1 signals first price, 2 second price and 3 an agreed deal price. The bidfloor attribute on each impression object carries the minimum the seller will accept, expressed as a cost per thousand impressions in the currency named by bidfloorcur. Most of the market still transacts on OpenRTB 2.5 and 2.6, according to IAB Spain's first supply-side platform guide of April 2026.

A worked example. A campaign values an impression at 8.00 dollars. The exchange declares at of 1 and a bidfloor of 2.00. The shading model, drawing on historical clearing prices for that publisher, ad slot, geography, device, time of day and exchange, predicts a minimum winning price near 3.20 dollars. The platform returns bid.price of 3.45, a shading factor of 0.43. The win notice resolves ${AUCTION_PRICE} to 3.45, and the surplus of 4.55 dollars stays with the advertiser rather than the publisher.

Bid prices in the response are usually net of the buying platform's fees, a detail recorded in the IAB Tech Lab Programmatic Auction Definitions published in final form on 26 June 2026. That document also confirms that few auctions clear on price alone: direct-sold campaigns, pacing, competitive separation and frequency caps can all outrank cost.

The feedback loop

Shading models require training data, and the industry built a channel to supply it. OpenRTB 2.5, released in 2016, added the lurl field, a loss notice URL the exchange calls when a bid is known to have lost. Bidders embed macros in it. The ${AUCTION_LOSS} macro returns a reason code, where 100 means the bid fell below the auction floor, 101 below a deal floor and 102 lost to a higher bid.

The commercially valuable macro is the minimum bid to win, implemented under names including ${AUCTION_MIN_TO_WIN} and ${AUCTION_MINIMUM_BID_TO_WIN}. Exchanges populate it only on competitive losses, code 102, and it reports the price the bid would have had to reach. Naming is not standardised: integration documents from Microsoft's Xandr, Digital Turbine, Opera and Verve each specify a different macro set. Google committed to the same disclosure in 2019, telling publishers that buyers would receive the minimum bid price needed to win after each Ad Manager auction closed.

Each loss therefore returns a labelled data point, and each win a clearing price. Over billions of auctions this produces a bid landscape, the estimated distribution of the highest competing bid for a given set of features. Published methods differ, from field-weighted factorization machines to deep distribution networks and nonparametric approaches, but resolve to one output: a win probability curve, and the bid on it that maximises expected surplus.

Who does the shading

Both sides of the market have built it. On the demand side it is standard in every large platform, including Display and Video 360, The Trade Desk, Amazon DSP and Yahoo. Google's documentation describes the behaviour without using the term: fixed cost-per-thousand bids in Display and Video 360 run through a setting called Optimized fixed bidding by default, which the help centre says aims to ensure buyers never overpay, and which must be unchecked for the stated bid to be submitted in full.

Sell-side shading appeared first and sits in tension with the seller's own interests. Exchanges launched tools that lowered buyer prices to keep volume from collapsing during the transition: AdX operated a Bid Translation Service, Rubicon Project introduced Estimated Market Rate and AppNexus offered Bid Price Optimization. Rubicon chief executive Michael Barrett described Estimated Market Rate to AdExchanger in November 2018 as a repurposed soft-floor product running the same logic in reverse: knowing an impression cleared at 3 dollars last time, and setting the bid at 3.10 to be sure of winning.

Origin and evolution

Header bidding, which let publishers solicit bids from several exchanges before calling the ad server, made second-price mechanics unworkable: a second-price winner entering a downstream comparison competed at an artificially deflated number. In January 2017 no major display exchange ran first-price auctions, according to the CIKM 2020 paper by Djordje Gligorijevic and six co-authors at Yahoo. AppNexus, Index Exchange and OpenX led the switch through 2017 and 2018. Google Ad Manager, the last major holdout, announced its move in March 2019 and began the full rollout on 10 September 2019, dropping the last-look advantage. By December 2019 the share of first-price auctions approached 100 percent, up from about 40 percent that January. AdSense followed in October 2021.

The price effects were immediate and disputed. Surveys cited by Yahoo researchers put the increase in average traffic prices between 5 and 50 percent, with one report finding that 10 percent of advertisers stopped bidding entirely after their exchange switched. The same team estimated that without shading, cost per thousand impressions on their platform would have roughly doubled. Vendor claims for early tools varied widely: AppNexus reported prices 25 percent lower over 100 days, Rubicon a 5 percent reduction over four months, and The Trade Desk said its Koa tool cut publisher rates by 20 percent. Barrett attributed fourth-quarter 2018 pressure on Rubicon's rates partly to such tools.

Why it matters for the marketing community

Shading determines what a fixed budget buys. A 500,000 dollar budget clearing at an average 4.00 dollar cost per thousand impressions delivers 125 million impressions; at 3.20 dollars it delivers roughly 156 million. The mechanism is also a fee surface. Infillion's April 2026 buyer guide lists bid shading alongside data onboarding and log-level access among items a demand-side platform may bill separately, warning that not all costs are immediately visible.

Limitations and disputes

The core criticism is that shading relocated an opacity problem rather than solving it. First-price auctions were promoted as removing the black box between a bid and a clearing price. Shading rebuilt one on the buy side, where an advertiser sees a final price without seeing the valuation, the predicted landing point, or the margin between them.

Yahoo's researchers made a sharper argument about the sell-side variants, noting that exchanges shading on behalf of buyers lower their own yield and therefore hold incentives that diverge from the advertiser's, and describing those services as useful for keeping impressions sold but not utility-maximising for buyers.

Publishers respond with floors, and the two systems now train against each other. Yahoo published results from a floor-setting model built explicitly to induce bidders to change behaviour, reporting annualised incremental revenue of 1.3 percent on display and 2.5 percent on video inventory.

Disclosure remains unsettled. The Media Rating Council's draft Digital Advertising Auction Transparency Standards, issued in September 2025, are the first formal attempt at auction disclosure rules; Check My Ads Institute filed a 15-page objection warning the framework risked legitimising opacity. The IAB Tech Lab's final auction definitions set out 15 terms including loss URL, received bids and bid price. Bid shading is not among them.

Adjacent terms that are not bid shading

Floor price is the seller's minimum, set in bidfloor. It constrains shading from below rather than performing it.

Bid deflation programmes run by intermediaries are separate. Project Poirot, described in Teads' August 2026 complaint against Google, deflated Display and Video 360 bids into non-Google exchanges while leaving bids into AdX untouched, with a second version raising deflation to as much as 90 percent. That is bid reduction applied selectively by a party with a stake in where the impression clears, not surplus optimisation for the advertiser.

Supply-path optimization selects which exchange to route through, and bid caching reuses a losing bid in a later auction. Both concern which auction a bid enters. Shading sets the number written into it.

Recent developments

Shading logic is migrating toward the inventory. Bedrock Platform became the first demand-side platform to run its bidder inside an exchange in April 2026, in cryptographically signed containers within Index Exchange infrastructure, removing the queries-per-second ceiling that forced buyers to evaluate only a sampled slice of the bid stream. OpenX's OpenXBuild suite applies partner algorithms inside the exchange on the same premise.

Agentic buying raises a new pricing question. A DataBeat report of 22 June 2026 found conventional demand clearing at 6.95 dollars against 6.13 dollars for agentic buyers, a 13.4 percent premium, with agentic systems participating in 86 percent fewer auctions. IAB Tech Lab shipped AAMP 2.3 on 30 July 2026 with a pricing provenance field intended to stop agents asserting prices without a traceable source.

Timeline

  • 1961 - William Vickrey publishes the analysis establishing shading as the rational strategy in sealed first-price auctions
  • 2016 - OpenRTB 2.5 adds the loss notice URL and loss reason codes, creating a standard channel for minimum-bid-to-win feedback
  • September 2017 - AppNexus, Index Exchange, OpenX, Rubicon Project and PubMatic begin first-price testing; Rubicon announces a dual auction model
  • 2018 - Sell-side pricing tools launch, including Rubicon's Estimated Market Rate and AppNexus's Bid Price Optimization
  • March 2019 - Google announces the transition of Google Ad Manager to a unified first-price auction
  • 10 September 2019 - Google begins the full first-price rollout across Ad Manager, ending last look
  • December 2019 - Share of display auctions running first price approaches 100 percent
  • October 2020 - Yahoo researchers publish machine-learning shading methods at CIKM
  • 7 October 2021 - Google announces the move of AdSense to first-price auctions
  • September 2025 - Media Rating Council issues draft Digital Advertising Auction Transparency Standards
  • 20 October 2025 - Check My Ads Institute files comments challenging the draft standards
  • 29 January 2026 - IAB Tech Lab opens its Programmatic Auction Definitions for public comment
  • April 2026 - Bedrock Platform runs a demand-side bidder inside Index Exchange infrastructure
  • 22 June 2026 - DataBeat reports conventional buyers clearing 13.4 percent above agentic demand
  • 26 June 2026 - IAB Tech Lab publishes the final Programmatic Auction Definitions
  • 30 July 2026 - AAMP 2.3 adds a pricing provenance field for agentic transactions

Summary

Who: Demand-side platforms including Display and Video 360, The Trade Desk, Amazon DSP and Yahoo run shading on behalf of advertisers and agencies. Exchanges including the former AppNexus, Rubicon Project and Google's AdX built sell-side equivalents. IAB Tech Lab maintains the OpenRTB fields the practice depends on, and the Media Rating Council has drafted the disclosure standards that would govern it.

What: The submission of a bid below an advertiser's private valuation in a first-price auction, calculated from a predicted distribution of competing bids and delivered through the bid.price field of an OpenRTB bid response.

When: The technique entered digital advertising with the first-price transition that began in 2017, reached near-universal application by December 2019, and remains standard as of August 2026.

Where: Inside the bidding logic of demand-side platforms and, in sell-side variants, inside exchanges. Since 2026 the same logic increasingly executes in containers running within exchange infrastructure.

Why: First-price auctions charge the winner exactly what was bid, so any gap between valuation and clearing price is lost surplus. Shading recovers that surplus for buyers, at direct cost to publisher rates, and it has become one of the least visible variables in programmatic pricing.