Ad Auction Mechanisms & Smart Bidding
Online ad auctions allocate ad slots to competing advertisers in real time (< 100 ms). The auction mechanism determines who wins, what they pay, and whether the equilibrium is truthful. This page covers the major mechanisms and the ML systems advertisers use to bid optimally.
Vickrey (Second-Price) Auction
The simplest truthful single-slot auction:
- advertisers submit sealed bids .
- The highest bidder wins: .
- The winner pays the second-highest bid: (the order-statistic notation for the second largest).
Dominant strategy: Bidding your true value is dominant. You can never gain by overbidding (win at a loss) or underbidding (lose an auction you'd have profited from). This makes the auction incentive-compatible.
Proof sketch: Suppose and the highest opponent bid is .
- Bid : win, pay 7, profit . ✅
- Bid (overbid): same win/loss decision if , no gain.
- Bid (underbid): lose if , leave 3 profit on the table. ❌
Generalized Second Price (GSP) Auction
Used by Google Ads, Bing Ads for multiple ranked ad slots:
- slots with click-through rates (slot 1 gets the most clicks).
- Advertisers are ranked by bid × predicted quality score (Google's quality score incorporates expected CTR, ad relevance, landing page quality).
- Advertiser ranked pays the next advertiser's bid (the bid that would just displace them from slot ).
where is the quality score of the advertiser in slot .
Not fully truthful (unlike Vickrey): the dominant strategy is not always to bid true value in GSP. But in practice, GSP converges to a Nash equilibrium where sophisticated advertisers bid approximately truthfully after learning.
VCG Mechanism
The Vickrey-Clarke-Groves mechanism generalizes second-price truthfulness to multiple slots:
Each winner pays:
VCG charges each advertiser their externality — the harm their presence causes to other advertisers by displacing them to lower slots. This is the theoretically optimal truthful mechanism for multi-slot auctions, but it's computationally complex and Google has documented that GSP approximates VCG equilibria in practice.
Reserve prices
A reserve price is the minimum eCPM required to show an ad:
Why set a reserve price?
- Revenue floor: prevents inventory from being sold below production cost.
- Quality control: low-bid, low-quality ads degrade user experience even if they "win."
- Incentive to bid higher: a reserve price near the second-bid equilibrium pushes advertisers to bid higher, increasing revenue.
ML-optimized reserves: platforms learn per-slot, per-advertiser reserve prices from historical auction data to maximize expected revenue (Myerson's optimal auction theory).
Smart Bidding: ML for advertisers
Smart bidding (Google Ads) / Advantage+ bidding (Meta) automates the advertiser's bid strategy:
Target CPA (Cost Per Acquisition)
The advertiser sets a target CPA. The bidding system finds the optimal bid per auction:
The system uses a conversion prediction model (trained on the advertiser's conversion signals) and adjusts the bid to match the target CPA over a rolling 7-day window.
Target ROAS (Return on Ad Spend)
Maximize conversion value (revenue) subject to a minimum ROAS constraint:
Portfolio bidding
Manage bids across a portfolio of campaigns, shifting budget toward campaigns with the highest marginal return:
where is the value function for campaign as a function of budget and is the total budget. This is solved with a Lagrangian approach, treating the shadow price of the budget constraint as the universal bid modifier.
Budget pacing
Without pacing, a daily budget exhausts in the first hour. Pacing adjusts the bid multiplier dynamically:
The pacing controller (typically a PI controller) computes by comparing actual spend rate vs. target spend rate:
where is the spend error. LinkedIn's system achieves < 5% budget under-delivery using this approach.
Related concepts
- Ad Ranking & CTR Prediction: the CTR model whose eCPM = CTR × bid feeds these auctions
- Diversity, Cold Start & Exploration: the budget/exploration concerns that sit alongside bidding in the ad stack