Paid Caching and Speed Plugins: When the Free Tier Stops Being Enough

A caching plugin that works is easy to mistake for a fast site. The page cache gets written, TTFB drops to 80 ms, the hosting dashboard turns green, and everyone moves on. Then the mobile report arrives with a 4.1-second LCP, because the theme loads nine scripts in the head and the hero image is a 1.8 MB PNG that nobody compressed.

That gap is where the paid tiers live. It is not page caching, which free plugins do well enough for most sites. It is the parts of the request the cache never touches: which scripts load on which template, when JavaScript is allowed to execute, and what the browser has to download before it can paint anything at all.

Every price and version below was read from the vendor's own page on 16 October 2026. Where a vendor does not publish a version, this article says so rather than guessing one.

Free caching solves the first request, not the expensive one

Page caching is a solved problem. A free plugin writes the rendered HTML to disk or to the object cache, serves it to the next visitor, and clears it when a post changes. If that is your whole performance problem, the free stack from our free speed plugin roundup is genuinely sufficient, and this article is not for you.

The problems that remain after page caching are browser-side, and they are specific. A contact page and a product page share a header script that only the product page needs. A chat widget loads before anything the visitor can read. An animation library parses for 300 ms so a carousel can fade in. On most sites the HTML arrives in under 200 ms and the browser then spends four seconds deciding what to do with it.

Measure that split before buying anything. The first command asks the server what it did; the second tells you how many plugins are competing for the same request.

curl -sI https://shop.example.com/product/example/ | grep -iE 'x-cache|cache-control|age'
wp plugin list --status=active --fields=name,version --format=table

If x-cache reports a hit and the browser still takes four seconds, a caching plugin is not the missing piece. That is the whole diagnosis, and it costs nothing.

What the paid licences actually add

WP Rocket, for the site with one or two templates and no full-time developer

WP Rocket is the option most people mean when they say "the paid caching plugin". It writes the page cache and, more usefully, manages the browser side: delaying JavaScript until interaction, removing unused CSS, preloading the cache after a publish, and serving images through its own CDN on a free allowance of three pages. The pricing page lists a 14-day trial, so the licence decision can wait until after you have seen your own numbers move.

Checked on 16 October 2026: the site identifies itself as WP Rocket 3.20.0.2, and the licence tiers are $59.95 a year for one site, $119.95 for three and $299.95 for fifty. A WP Rocket licence buys updates and support for the period, so this is a subscription, not a one-off purchase.

It stops being enough in two places. It cannot make a slow query faster, because the cache stores the result of that query rather than replacing it — the object cache layers in our object cache invalidation notes are what you need there. And on a site that renders different HTML for logged-in users, full page caching is the wrong tool no matter who sells it.

Perfmatters, for the site whose theme loads scripts nobody uses

Perfmatters does not write a page cache at all. It is a control panel for everything else: disable a plugin's scripts on the pages that do not need them, delay JavaScript, remove unused CSS, host Google Analytics locally, and clean the database tables that accumulate revisions and transients. It pairs with a caching plugin rather than replacing one, which is why the two appear together in most speed stacks.

Checked on 16 October 2026: the changelog is at 2.6.9, dated 10 June 2026, and the pricing page lists $29.95 a year for one site, $59.95 for three and $124.95 for unlimited sites. Renewals get a 15% discount applied automatically. A Perfmatters licence is the cheapest way to find out whether your problem is scripts; if the score does not move after two hours of exclusions, your problem is not scripts.

Where it stops being enough: it gives you switches, not decisions. Every exclusion is a manual judgement that has to be retested after the next plugin update, and on a site with a marketing team adding tools monthly, that maintenance is the real cost. It also does nothing for images, which is usually the larger number on a mobile LCP.

NitroPack, for the owner who wants the decisions made elsewhere

NitroPack optimises from outside the site. Its service holds a copy of the page, rewrites it at the edge, and hands back something already compressed, delayed and lazily loaded. There is a free plan, which makes it testable, and the paid tiers are priced by traffic rather than by site count.

Checked on 16 October 2026: the pricing page lists the entry tier at $22 a month, or $18 a month billed annually, and the high-traffic tier at $99.99 a month or $83.01 billed annually, with enterprise pricing starting at $229 a month. A NitroPack plan is worth reading before you compare it with a plugin, because you are buying a service rather than a file you install.

Where it stops being enough: the configuration lives in somebody else's dashboard. When an over-eager delay breaks a checkout script, you debug inside their interface, and you cannot carry the settings to another host. For a site with no staging environment and no developer on retainer, that trade is often the right one. For a store with a release process, it is an extra layer of opacity.

The paid upgrade that changes the ceiling is hosting

Once the HTML is cached and the scripts are trimmed, the wins left are object caching, PHP workers and network distance. Those are hosting decisions, and they are the ones people arrive at after buying two plugins that did not help.

Rocket.net, when the origin is the queue

Rocket.net puts a full-page cache at the edge in front of every site and isolates PHP workers, which is what stops one slow plugin from holding up everybody else's request. Checked on 16 October 2026: plans start at $30 a month, or $25 a month billed annually, with a $1 first month on the entry plan. A Rocket.net plan is a reasonable first stop for a store where the plugin stack is already tuned and the server is still the slow part.

Kinsta, when you want the cache and the database in one place

Kinsta runs its own page cache and object cache and manages the database layer, so the caching plugin conversation changes shape: the host is already answering the first request. Checked on 16 October 2026: the Starter plan is $35 a month for one WordPress install, with a discount for paying annually. A Kinsta plan suits a site that will not have a developer reviewing plugin exclusions every month.

Hosting has its own limit, and it is worth stating plainly. Moving to faster infrastructure does not fix a large image, an unremoved render-blocking stylesheet, or a 4 MB hero. Those belong to the image pipeline, and no host will do them for you.

Who should stay on the free stack

If your site has fewer than twenty pages, no logged-in users, no checkout, and no third-party tools beyond analytics, paid speed plugins are a solution looking for a problem. A free caching plugin, an image pipeline that actually outputs WebP at the right dimensions, and a free CDN tier will get you most of the way for nothing, and the free roundup covers exactly that combination.

The paid licences start making sense when the free tier has a specific, reproducible failure: you need to disable a script on one template and the free plugin only offers a global switch; you need to exclude a URL from the cache and the plugin ignores the query string; you need a preload after publish and your host does not provide one. Buy on that failure, not on a score.

One more honest limit. None of this makes a slow site fast if the bottleneck is a query that scans a million rows, and none of it substitutes for a CDN when your audience is on another continent. Plugin licences buy control and convenience. They do not buy an architecture.

What to do on Monday

1. Measure TTFB and LCP separately on one template group. If TTFB is fast and LCP is not, stop looking at caching plugins.
2. Fix images before buying anything. That is the largest single number on most mobile reports.
3. If the report says scripts, take the 14-day WP Rocket trial or the Perfmatters licence, apply the obvious exclusions, and re-measure on the same template and the same device profile.
4. Only if both are clean and the site is still slow at the origin, price the hosting move.
5. Keep a note of the version you installed. A speed stack is a configuration, and nobody remembers which exclusion they added eleven months ago.