How Amazon Teams Think About Catalog Latency During High-Intent Traffic

Catalog latency problems are rarely caused by one obvious bottleneck. More often, the issue is cumulative: search enrichment, pricing checks, recommendation joins, and image metadata all add cost to a path that the user experiences as one page load.

A disciplined team starts by defining the page elements that must be fresh versus the ones that can tolerate soft staleness. That boundary determines which data can be cached aggressively and which dependencies need tighter budgets.

The operational trick is to degrade gracefully before the page fails entirely. If a non-critical dependency times out, the user should still see a fast page with fewer enrichments rather than a broken request chain.