The Economics of Being Chosen: An Introduction to Auction Theory for Local-Service Advertisers

A rate card names a price. An auction discovers one.

A local business owner reasonably expects advertising to have a price list. You choose a keyword, you see a cost, you decide whether it is worth paying. Sponsored search does not work that way. The amount a business pays for a click is not posted in advance by the platform. It is produced, in real time, by an auction that runs the instant a person searches, and it settles at a figure that depends on who else wanted the same query at the same moment.

This is not a minor accounting detail. It is the core fact that separates managing paid media from merely funding it. Benjamin Edelman, Michael Ostrovsky, and Michael Schwarz set out the mechanism in the American Economic Review in 2007, calling it the generalized second-price auction and noting that it was already, at the time, selling billions of dollars of keywords a year. Everything that follows in this article is an attempt to take that one peer-reviewed finding seriously.

How the generalized second-price auction works

In a generalized second-price auction, advertisers submit bids for a search term. The highest bidder wins the most prominent ad position, the second-highest wins the next position, and so on down the page. The defining feature is in the pricing rule: a winner does not pay their own bid. Each advertiser pays a price derived from the bid of the competitor ranked immediately below them. You are, in effect, charged close to the minimum you would have needed to bid to hold your position, not the maximum you were willing to pay.

The design is a generalization of the classic single-item second-price auction studied by William Vickrey, extended to a page with several ranked slots rather than one prize. Because there are multiple positions of unequal value, the pricing and incentives become considerably more intricate than the single-item case, a point Hal Varian formalized in his 2007 analysis of position auctions. The mechanism is elegant, widely deployed, and, as the same authors showed, far less simple to bid into than it first appears.

Why a "just bid higher" google ads bidding strategy is naive

The intuitive assumption about a second-price auction is that it is safe to bid your true value, because you will only ever pay a shade above the next competitor. In the single-item Vickrey auction that intuition is correct and provable: truthful bidding is a dominant strategy. The generalized second-price auction breaks that guarantee. Edelman, Ostrovsky, and Schwarz proved that despite its resemblance to a Vickrey auction, GSP is not incentive-compatible: it generally has no dominant-strategy equilibrium, and bidding one's true value is not, in general, an equilibrium.

The practical translation is direct. There is no single bid that is correct regardless of what everyone else does. The right amount to bid for a query depends on the bids and behavior of the other advertisers competing for it, and those change continuously. This is why the advice to "just bid higher" is analytically wrong, not merely aggressive: it treats a strategic, competitor-dependent problem as a mechanical one. A defensible google ads bidding strategy is the ongoing work of modeling that competition, not a number chosen once and left alone.

The quality weight: why the highest bid does not always win

A further complication sits on top of the auction. In the position-auction models that describe real sponsored search, advertisers are not ranked by their raw bid alone. Each bid is weighted by an estimate of how likely the ad is to be clicked, so that expected value, rather than headline willingness to pay, helps determine rank and price. Varian's position-auction framework and the Edelman, Ostrovsky, and Schwarz analysis both build this click-weighting into the equilibrium.

For an advertiser, the consequence is that relevance is a lever on price, not only a matter of taste. A more relevant ad and a matched landing experience can hold a strong position at a lower bid than a less relevant competitor, because the auction is ranking a quality-adjusted bid rather than a raw one. This is the formal reason that improving the ad itself, and not only raising the bid, changes what a business pays for a given slot.

The mechanism the industry did not choose

If GSP is not truthful, an obvious question follows: why not use an auction that is? One exists. The Vickrey-Clarke-Groves mechanism, built from the work of William Vickrey, Edward Clarke, and Theodore Groves, charges each winning bidder the cost, or externality, that their presence imposes on everyone else in the auction. Under VCG, bidding your true value becomes a dominant strategy again, and the design carries attractive efficiency properties, including under bidder collusion.

The advertising industry adopted GSP anyway. The standard account is that GSP was simpler to explain to advertisers and, under many conditions, produced higher expected revenue for the auctioneer, which is enough to explain the choice. The point for a practitioner is not that the platforms picked the wrong auction. It is that they picked a specific auction with known, documented properties, and that those properties, not a rate card, govern what a campaign costs. The market you transact in was designed, and it does not reward naive bidding.

What this means for a local-service advertiser

Local-service categories, a med-spa, a plumber, a dental practice, a small law firm, sit inside this mechanism like everyone else, often with the added pressure of a small field of competitors bidding hard on the same high-intent local terms. Three consequences follow directly from the theory, without needing any vertical-specific figure to be true.

• Your cost per click is set by rivals, not by you. Because pricing is derived from competing bids, an increase in what a competitor is willing to pay raises your price even if your own strategy has not changed.
• There is no correct bid to set and forget. Since GSP has no dominant strategy, the appropriate bid moves as the field moves; a bid that was efficient last quarter can be wasteful this one.
• Relevance is a price lever. Because rank reflects a quality-adjusted bid, a sharper ad and a matched landing page can lower the price of a given position, which is work that a raw bid increase cannot do.

Reading the evidence

The claims in this article rest on established, peer-reviewed economics. The generalized second-price auction, its lack of a dominant strategy, and the truthful VCG alternative are documented in the American Economic Review, the International Journal of Industrial Organization, and the founding mechanism-design papers, not in vendor marketing. That is the strongest tier of evidence, and it is why the mechanics above can be stated plainly rather than hedged.

Two boundaries are worth marking. First, auction theory describes how prices form; it does not, by itself, tell you whether a given campaign is profitable. That is a separate empirical question about incrementality, and the field-experiment literature on it is a subject for its own article. Second, the specific price of a click in your market is not something a paper can supply. It is a measured quantity, and any number offered without a measurement behind it deserves suspicion, including numbers offered by an agency.