Bid shading

Bid shading is the practice, by a DSP or its bidding model, of bidding below the estimated value of an impression in a first-price auction so as to pay closer to the price needed to win rather than the full value.

Under second-price rules a bidder could bid its true value and let the auction discount it. Under first-price rules the bidder must do the discounting itself, by predicting the clearing price from past auctions on similar inventory and bidding a margin above it. The prediction improves with feedback, which is why OpenRTB 2.6 defines ${AUCTION_MIN_TO_WIN}: "minimum bid to win the exchange's auction," replaced with an empty string when the exchange withholds it and otherwise with "the minimum price needed to tie the winning or next-closest bid" (OpenRTB 2.6, section 4.4.1). The 2.6-202606 snapshot of 11 June 2026 added ${AUCTION_DISCOUNT_PCT} and ${AUCTION_DISCOUNT_CPM} so a seller can communicate a discount applied after the auction, and clarified that ${AUCTION_PRICE} is the final price after discount (openrtb2.x releases).

Bid shading is sold by DSPs as a buyer service and sometimes charged for, which makes it one of the fees inside the gap between what an advertiser spends and what a publisher receives. It is also one of the model behaviours that an explainable-bidding audit has to be able to reproduce: for a given request, what value did the model estimate, what clearing price did it predict, and why was the bid where it was. The simulator in Explainable AI in RTB shows the three numbers side by side.

Example: a DSP values an impression at $3.00, predicts a clearing price of $1.90 from recent ${AUCTION_MIN_TO_WIN} feedback on the same placement, and bids $2.05.

Related: first-price auction, floor price, DSP, working media.

Source: IAB Tech Lab, OpenRTB 2.6, section 4.4.1; openrtb2.x release 2.6-202606.

1. OpenRTB 2.6, section 4.4.1
2. openrtb2.x releases