User-perceived latency and responsiveness — Core Web Vitals, bundle size, rendering path, network waterfall, and request-path DB efficiency.
The user-facing speed layer: does the user wait longer than they should?
In scope. Core Web Vitals (LCP, INP, CLS), time-to-first-byte, bundle size and code-splitting, render-blocking resources, caching of user-facing responses, image optimization (format, dimensions, loading), database-query latency that shows up in request time, perceived-responsiveness patterns, streaming, prefetch / preload hints.
Not in scope. Server-side resource consumption without user-visible latency — that's cost-efficiency. Cache-correctness / staleness bugs — that's data-integrity. Rendering defects that break assistive technology — that's accessibility.
Distinct because. The defect is user wait. cost-efficiency's defect is meter runs unbounded, regardless of whether a user notices. A pattern about "hero image missing preload" is performance. A pattern about "handler queries unbounded rows per request" is cost-efficiency primary, performance secondary (it also hurts latency but the primary harm is resource burn).
Conceptual sub-structure. Loading performance, interaction responsiveness, visual stability, request-path efficiency.