Bid shading
Bidding less than you think it's worth, on purpose.
What it means
The strategy of submitting a bid below one's true valuation or raw estimate to improve the expected profit from winning. In first-price and common-value auctions it is rational: bidding your full value would leave no surplus, and in common-value settings winning itself is bad news, so shading corrects for the winner's curse. How much to shade depends on the auction format, the degree of value uncertainty, and the number of rivals — but the direction of that last effect flips by format. In an independent private-value first-price auction, more competitors force you to shade less and bid nearer your value; in a common-value auction, more competitors worsen the winner's curse, so you shade more. Correct shading is the practical antidote to overpaying, and failures to shade, by inexperienced bidders, are a documented source of losses.
The private-value case
Shading resolves a trade-off. Bid closer to your value and you win more often but keep less surplus; bid lower and each win is more profitable but rarer. The optimum balances the two. In the textbook symmetric case — bidders drawing independent private values from a uniform distribution and competing in a sealed first-price auction — the equilibrium bid is your value scaled by (n-1)/n, so you shade by roughly your value divided by the number of bidders. With two rivals you hold back about a third of your value; with ten, only a sliver. The logic is forward-looking: you bid as if pivotal, asking what you would need to beat the highest competing value, not what the item is worth to you in isolation.
The winner's curse counterweight
When bidders share a common but uncertain value — an oil tract, a company, a spectrum licence — a second reason to shade appears. Winning means your estimate topped everyone else's, which is evidence it was too high. The three petroleum engineers who named the winner's curse in 1971 argued that oil firms bidding their honest estimates earned poor returns "year after year" on offshore leases. The remedy is to shade for adverse selection, and the correction grows with the field: more rivals make winning worse news, not better. Laboratory auctions confirm the difficulty. Kagel and Levin found subjects routinely overbid and lost money, and, tellingly, bid more aggressively as the number of competitors rose — exactly the wrong direction.
Bid shading in ad auctions
The idea went mainstream when programmatic advertising abandoned the second-price auction. As header bidding spread and exchanges — Google's ad server among them — moved to first-price clearing around 2018 and 2019, buyers who kept bidding their full valuation simply overpaid on every impression. Demand-side platforms responded with automated bid shading: predict the lowest price that still wins, then bid just above it. The hard part is that the minimum winning price is censored — you never observe it, whether you win or lose — so there is no direct label to learn from. Recent systems from Yahoo's ad team model the whole distribution of the winning price and optimise expected surplus against it, shading millions of bids per second.
Limits and caveats
Shading only pays if your model of the competition is right. Overestimate how low you can go and you lose auctions you would have profited from; underestimate and you hand back the surplus. It also depends on the format: in a true second-price or Vickrey auction, bidding your value is the dominant strategy and shading is a mistake, because the price you pay is set by someone else, not by your own bid. And shading is inherently unstable as a collective act — when every buyer trims, clearing prices drift down and each bidder's best response shifts again, which is one reason first-price ad markets needed floors and repeated recalibration rather than settling into a tidy equilibrium.
Examples
In a sealed first-price auction with few rivals, a buyer who values an item at $100 bids only $80 to leave room for profit.
Ad exchanges that moved from second-price to first-price auctions now shade bids automatically on advertisers' behalf, because paying your full valuation on every impression leaves no margin at all.
A firm bidding to acquire a startup trims its offer below its own valuation: when five rivals are guessing at the same uncertain future revenues, winning means you guessed highest.
At a sealed-tender government bond auction, a primary dealer offers a price below its own valuation, shading so that winning the allocation still leaves a margin to resell in the secondary market.
A construction firm bidding for a public contract prices above its true cost, then trims that markup toward cost as more rivals enter, since extra competition rewards a leaner bid.
First described in Auction theory; Vickrey (1961); applied to the winner's curse.
Key references
- Zhou, T., He, H., Pan, S., Karlsson, N., Shetty, B., Kitts, B., Gligorijevic, D., Gultekin, S., Mao, T., Pan, J., Zhang, J., & Flores, A. (2021). An efficient deep distribution network for bid shading in first-price auctions. Proceedings of the 27th ACM SIGKDD Conference on Knowledge Discovery & Data Mining (KDD '21). doi.org/10.1145/3447548.3467167
- Gligorijevic, D., Zhou, T., Shetty, B., Kitts, B., Pan, S., Pan, J., & Flores, A. (2020). Bid shading in the brave new world of first-price auctions. Proceedings of the 29th ACM International Conference on Information & Knowledge Management (CIKM '20), 2453-2460. doi.org/10.1145/3340531.3412689
- Kagel, J. H., & Levin, D. (1986). The winner's curse and public information in common value auctions. American Economic Review, 76(5), 894-920. ideas.repec.org/a/aea/aecrev/v76y1986i5p894-920.html
- Riley, J. G., & Samuelson, W. F. (1981). Optimal auctions. American Economic Review, 71(3), 381-392. ideas.repec.org/a/aea/aecrev/v71y1981i3p381-92.html
- Capen, E. C., Clapp, R. V., & Campbell, W. M. (1971). Competitive bidding in high-risk situations. Journal of Petroleum Technology, 23(6), 641-653. doi.org/10.2118/2993-PA
- Vickrey, W. (1961). Counterspeculation, auctions, and competitive sealed tenders. The Journal of Finance, 16(1), 8-37. doi.org/10.1111/j.1540-6261.1961.tb02789.x