Download websites that use Performance Lab

About Performance Lab

Most plugins are installed because somebody wanted a feature. Performance Lab is installed because somebody wanted to test one. The plugin exists as an opt-in channel for experimental work by the WordPress.org performance team, features that are candidates for the core platform but not yet committed to it, so every installation represents an operator who chose to run unfinished code on a production site on purpose.

The choice is the signal, not the software. As of October 2026 the plugin appears on 1,829 sites, a modest population against the Performance ecosystem it belongs to, and a population selected for a particular temperament. Operators who install an experimental performance plugin read release notes, measure their own pages and are willing to be early. Sites in this list are more likely than average to be professionally maintained, and much more likely to accept a new tool if it can demonstrate a measurable improvement.

Release cadence makes the version data unusually informative. The plugin ships often, and across the install base we can currently distinguish 19 release strings, of which 2.7.0 is the one we record most often. On software whose entire purpose is early access, a stale version is a contradiction, so the tail of old releases is really a list of sites that opted in once and then stopped paying attention. The freshly updated rows are the genuinely active ones, and the gap between the two groups is the part of the list worth acting on.

The technical context bounds what any of this proves. Written in PHP, distributed under the GPL v2 licence through the official plugin repository, it costs nothing to install and nothing to abandon, so presence measures interest rather than commitment. Detection depends on the plugin's assets or markers surviving into the served page, which caches and asset optimisers can hide, so the count is conservative, and the bias again works against the most aggressively optimised sites, which are the ones likeliest to be running it.

For a researcher the value of the export is that it identifies a behaviour rather than a purchase. Anyone building performance tooling, hosting, image services or monitoring wants operators who already measure their own sites, and no directory lists those people. Narrowing a population of publishing-platform sites down to the ones that opted into experimental performance work is a qualification step a per-domain lookup cannot perform at all, because the question is not what a single site runs but which sites, across the whole population, behave a particular way.