Performance work goes wrong when teams optimize whatever is easiest to
graph. For commerce, start with real user monitoring or lab-based
journeys: time to first byte on category and product pages, add-to-cart
latency, and checkout steps. Pair that with server-side signals: PHP-FPM
pool saturation, slow query logs, Redis hit rates, and queue health.
Profiling matters, but so does discipline: fix the top one or two
bottlenecks that align with revenue paths before chasing micro-optimizations.
Full-page cache and Varnish help public pages; authenticated and checkout
flows often need different strategies.
Document baselines before and after changes. That single habit turns
“we made it faster” into something you can defend in a retrospective—or
roll back with confidence when a deploy does not land as expected.