The Fastest SSP Decision Is the One That Has Already Been Made

In the world of digital advertising, auction systems must make quick decisions to determine which demand partners should bid for a publisher's ad inventory. These auctions happen rapidly, presenting a challenge for engineers seeking to streamline the process. Instead of continuously optimizing every aspect of the auction path, a more effective approach is to identify and eliminate unnecessary work.

The ideal auction system should be predictable, with a minimal number of dependencies, and capable of handling scaling traffic. By separating the responsibilities of management and execution, a more efficient architecture can be achieved. Authoring, which involves creating business intent, rules, and configurations, can be designed for flexibility and validation. In contrast, execution should focus on speed, stability, predictable latency, and simple failure behavior.

Pre-computation can further enhance performance by preparing responses or parts of them before the request arrives. This can include transforming eligibility rules into cheap evaluation structures, materializing partner and publisher settings closer to the serving path, and validating complex configurations before they are used in live traffic.

However, this approach requires careful consideration of freshness vs. latency, as pre-computed data may have a propagation delay between when a change is made and when it becomes visible to the execution plane.

Ultimately, the goal is to minimize unnecessary work in the auction path while ensuring the freshness of data is maintained according to its specific requirements. By focusing on this separation of concerns and pre-computation, the auction system can operate more efficiently, resulting in faster response times and reduced infrastructure costs.