Computational Advertising vs Recommendation Systems: Key Differences and Popular Models

Computational advertising and recommendation systems both aim to predict user interactions, but they solve different core problems: advertising coordinates the interests of advertisers, users, and media, while recommendation focuses on enhancing user experience.

Because of these differing goals, the two fields adopt distinct modeling strategies. In advertising, CTR models are used primarily to maximize revenue, often ranking high‑CTR ads at the top, whereas recommendation systems balance CTR with diversity, novelty, and long‑term user satisfaction.

Advertising also involves many auxiliary modules such as bidding strategies, budget control, and ad scheduling, which rely on game theory, optimization, and control theory—areas typically absent from recommendation pipelines.

Key technical topics for each domain are summarized as follows:

Computational Advertising: CTR modeling, bidding strategies, yield optimization, intelligent budget control.

Recommendation Systems: CTR and other ranking models, exploration‑exploitation trade‑offs, cold‑start problems, data bias handling.

The article then lists the mainstream models used in industry advertising, categorized into traditional machine‑learning models (LR, FM/FFM, GBDT+LR) and deep‑learning models such as pure DNNs, dual‑tower architectures (DSSM, variants), Wide & Deep families (Wide & Deep, DeepFM), attention‑based networks (DIN), and several emerging approaches (DCN, DIEN, Transformer‑based models, reinforcement‑learning methods).

Among these, the author highlights three models that are widely adopted across different business scales: GBDT+LR for small‑to‑medium workloads, DSSM for semantic matching, and Wide & Deep for a balanced combination of memorization and generalization.

Finally, the author reflects on the career of a computational‑advertising or recommendation‑system algorithm engineer, noting the fast‑changing nature of the field and encouraging readers to share their own experiences.