Generalized Second-Price (GSP)

The generalized second-price (GSP) auction is the multi-slot mechanism search advertising made famous: bidders rank for ad positions, and each winner pays just enough to beat the bidder below — the 'second price,' generalized across several slots at once. It is the model the canonical literature analyzed (Edelman, Ostrovsky and Schwarz's 2007 AER paper; Hal Varian's position-auctions work) and the mental model a generation of search marketers bid against — with the modern caveat that the live auction has accumulated documented modifications the textbook never had.

The clean theory first: slots descend in value, bidders are ranked (in practice by bid times quality — ad rank, which is why QUALITY-SCORE folklore exists), and each pays the minimum needed to hold their position. GSP's famous theoretical wrinkle: unlike the single-item VICKREY auction it resembles, GSP is NOT truthful — bidding your true value is not always optimal, and the equilibrium-bidding literature exists precisely because rational bidders shade strategically even here (a milder cousin of the FIRST-PRICE era's BID-SHADING science). The modern reality carries the asterisks the antitrust era put on the record: Google's search auction has run RGSP — randomized GSP — since 2019, occasionally flipping adjacent ranks (with the runner-up's price rising toward parity), a mechanism trial documents described internally as 'a better pricing knob'; 'squashing' adjusted predicted-CTR inputs with broad price effects; and launches like Project Momiji tuned the auction toward revenue targets — the DOJ record establishing that clearing prices respond to house levers, not bids alone. The practitioner translation is not cynicism but posture: treat the search auction as a goal-driven black box (the SMART-BIDDING era's strategies already do — you state economics, the system bids), keep the GSP model for intuition about why position prices step the way they do, and price the meta-lesson alongside the display market's: auction formats are vendor products, and their physics can be revised.

GSP literacy matters for reading search economics — why position upgrades cost step-functions, why quality multiplies money, why 'pay the next bid' was never quite 'pay your value' — and for interpreting the era's disclosures without panic or naivety. It matters most as posture: bid through goal-based strategies against measured value (the allowable-CPA math this glossary's economics entries build), audit outcomes rather than trusting format theory, and remember both halves of the decade's auction story — display's format flipped openly in 2019; search's was revised quietly and litigated later.

GSP grew organically inside early search advertising (Google generalized Overture's designs), was canonized by the 2007 Edelman-Ostrovsky-Schwarz and Varian papers, and acquired its asterisks in public: RGSP from 2019, and the 'pricing knob' internal language the DOJ trial put on the record — the textbook auction, annotated by its own emails.

Etymology: source.

Search interest for this term over the last five years:

View interest-over-time on Google Trends →

Areas of marketing where generalized second-price (gsp) is a core concern: