Researchers solve capacity-constrained ad auction problem with first truthful mechanism
First constant-approximation truthful mechanism for ads of different sizes under a global space limit.
Traditional position auctions assign fixed slots to advertisers to maximize welfare, a standard matching problem. But modern ad formats allow varying ad sizes (images, videos) that compete for limited screen space rather than slot slots. Batziou et al. formalize this as a matching problem with a global capacity constraint: the platform selects a subset of ads and positions them under a total size budget. Greedy methods fail to guarantee near-optimal welfare.
The authors develop an algorithm that first orders ads by density (value per unit size) and then applies capacity-aware local improvements—swapping or moving ads to better positions without violating the constraint. This yields a constant-factor approximation guarantee. By adding a small randomization step, they create a universally truthful mechanism (advertisers bid truthfully regardless of randomness). Their work is the first to achieve both truthfulness and a constant approximation for this practical yet theoretically challenging setting.
- Proposes a density-based ordering algorithm augmented with local re-allocations for capacity-constrained ad matching
- Achieves constant factor approximation guarantee for welfare maximization under a global space constraint
- First universally truthful randomized mechanism with constant approximation for this variant of position auctions
Why It Matters
Enables ad platforms to optimize revenue and user experience when ads have different sizes, under a physical space limit.