A caching plugin stores a finished copy of your page and serves that copy instead of rebuilding it on every request. It genuinely helps, and it also lets a slow site produce a good score on a speed test without being faster for the person waiting on it. Google measures the person.
What caching actually solves
Without caching, every visit makes WordPress run PHP, query the database, assemble the page and send it. That work takes time, typically a few hundred milliseconds on decent hosting and far more on cheap shared hosting.
Caching removes that work for repeat requests. It is real, it is worth having, and every serious host either includes it or expects you to. The improvement lands almost entirely in one number: time to first byte, meaning how long before the server starts responding.
What it does not solve
A cached page is the same page. If it contains three megabytes of CSS and JavaScript from a page builder, the cached version contains the same three megabytes.
Caching does nothing for page weight, nothing for the number of requests, nothing for oversized images, nothing for render blocking scripts, and nothing for layout shift. Those are properties of what the page is, not of how quickly the server produced it.
So the visitor on a phone on a mediocre connection, who is the visitor Google is measuring, sees a small improvement at the start and then waits exactly as long as before for everything else.
Why the test score improves anyway
Testing tools measure from a fast connection in a data center. Server response time is a visible share of a fast test and a small share of a real one. Improve the fast part, and the test result moves more than the real experience does.
This is the gap between lab data and field data, and it runs in one direction. Field data in Search Console is what Google uses, and it comes from actual visits on actual devices. When a site owner tells me their speed is fine because a tool said so, this is almost always what happened.
Where caching plugins make things worse
Most of them do more than cache. They minify, combine, defer, lazy load and inline. Those features are genuinely useful and they are also where sites break.
Combining JavaScript files can change execution order and break a form. Deferring the wrong stylesheet produces a flash of unstyled content. Aggressive lazy loading applied to a hero image makes Largest Contentful Paint worse, because the single most important image is now loaded last. That last one is common and counterintuitive, and it means a plugin installed to improve speed measurably damaged the metric that matters most.
There is also the plugin’s own weight, and the fact that a stack of optimization plugins is itself a maintenance liability.
The order that actually works
- Fix the page weight. Size and convert images properly. Remove scripts nothing uses. Identify the plugins responsible for the bulk of it.
- Fix layout shift. Set dimensions on images and embeds. Cheap and immediately noticeable.
- Then cache, ideally at the host level rather than through a plugin.
- Measure in the field, not in a tool, and confirm real visitors improved.
Doing this in reverse is how sites end up with four optimization plugins, a score of 95, and visitors who still wait four seconds.
What this site does
palmprojects.com uses the caching layer its host provides and no optimization plugin at all. There is nothing to minify, because the site ships one stylesheet and one small script. There is nothing to combine, because there are only two files. There is nothing to defer, because nothing blocks rendering.
That is the point worth taking. Optimization plugins exist to compensate for pages that were assembled badly. A page that was not assembled badly does not need them, and the numbers are published rather than claimed.
Questions
Should I remove my caching plugin?
Not necessarily. If your host does not provide caching, keep it. Consider turning off the minify and combine features and testing whether anything actually got worse. Frequently nothing does.
My host says caching is built in. Do I need a plugin too?
Usually not, and running both causes conflicts. WP Engine, Kinsta and similar hosts handle it at the server level, which is better than doing it in PHP.
Why does my score change between tests?
Lab tests vary with network conditions and server load. A single test tells you very little. Run several, and weight the field data above all of them.
