Can performance debugging be limited?

Yes. When you turn on the front-office profiler in MPR Performance Revolution, you decide who can see it: leave the panel off entirely, or show it but restrict it to logged-in employees so customers never see your timings.

One of the most useful PrestaShop performance tips is to profile the live store, not a clean test copy, real traffic, real modules, real data. The profiler adds a small bar that shows each page's SQL query count and time, and lists the slowest queries and slowest module hooks, so you can see exactly where a page spends its time before you change anything.

The common mistake is leaving a profiler visible to the public: it leaks internal details and adds overhead on every request. Keep it employees-only, switch it off once you've found the bottleneck, and act on what it shows, usually a heavy module hook or an unindexed query. The module also handles Redis object caching, asset deferral and OPcache/Smarty cache warming, so the fixes live in the same place as the diagnosis.

The relevant settings are two switches: profiler_panel_enabled turns on the front-office panel, and profiler_employees_only keeps it behind an active back-office employee session. The employee-only setting defaults to on, while the panel itself defaults to off. When profiling starts, the module initializes instrumentation early in the request and then injects the panel before </body> only if the collector is enabled and the employee gate passes.

For deeper SQL and hook attribution, the module includes optional overrides that record every database query and per-module hook timing, and the query profiler logs slow queries over the configured threshold, which defaults to 0.1 seconds. Install those only when you need that level of visibility, because they deliberately add instrumentation overhead. A good workflow is: enable employees-only profiler, reproduce one slow URL, capture query/hook totals, disable it, then make the smallest cache, index or module change that addresses the measured bottleneck.