performance-engineer

Start by measuring a representative baseline, then use profiling or distributed tracing to find the dominant bottleneck. Change one variable at a time and re-run the same benchmark to confirm the improvement is real.

Measure current Core Web Vitals values first, then address the largest contributor such as render-blocking resources, oversized assets, or slow server responses. Verify each change against the same metrics to avoid regressions.

Inspect the query execution plan to find full scans or missing indexes, then apply a targeted fix and re-measure. Avoid changing schema or queries based on intuition without baseline timing evidence.

Run load tests to validate expected capacity, stress tests to find breaking points, and soak tests to detect memory leaks or degradation over time. Always test against a defined performance objective tied to user-visible impact.

Optimizations fail when applied without measurement, when they shift the bottleneck elsewhere, or when they add caching and concurrency complexity with hidden trade-offs. Re-benchmarking after each single change prevents this.

Synthetic benchmarks may not reflect real traffic patterns, data distributions, or infrastructure constraints. Results should be labeled as assumptions until verified against representative production-like conditions.