Website performance
What should I measure before upgrading a hosting plan?
Short answer
Measure where time and resources go: server response time on uncached pages, CPU and memory under load, database query time and external calls. Upgrade only after you know which resource is the bottleneck.
What changes the answer
- Time to first byte for uncached requests
- CPU, memory and swap during busy periods
- Slow database queries and slow external services
A bigger plan fixes only what it increases
More CPU helps when the processor is saturated. More memory helps when the server is running out of it. Neither helps if a page is slow because of one database query, a call to a slow external service, or a large unoptimised image. Paying for resources that are not the bottleneck changes your bill and not your site.
Separate the server from the browser
Time to first byte (TTFB), described on web.dev, covers the time from the request until the first byte of the response arrives. It includes the work your server does. Rendering time in the browser comes after that. If TTFB is low and the page still feels slow, the server is probably not the problem.
What to record
- Server response time for representative pages, with caching both on and off, since cached results can hide a slow backend.
- CPU load, memory use and swap activity during your busiest periods.
- The slowest database queries, using your database's slow-query logging.
- The time taken by calls to external APIs.
- Whether processes were killed or requests queued.
Write the numbers down with the date and the conditions so you can compare after a change.
Interpret before buying
If memory is exhausted or CPU stays saturated at normal load, a larger plan may help. If requests queue while CPU and memory are idle, look at worker limits or a blocking call. WordPress's optimization documentation is a useful place to start for a WordPress site. HostCritiq does not publish provider benchmark results, so use measurements from your own workload.
Make one change at a time
Change one thing, measure again and keep notes. That tells you whether the upgrade or something else made the difference.
Keep a simple log
A short table is enough: date, page tested, caching state, response time, CPU and memory observed, and what changed since last time. After an upgrade, repeat the same tests under similar conditions. If nothing improved, you have learned that the plan was not the bottleneck before paying for it a second time.