Four clusters, every entry
Every theme and plugin in the directory follows the same four-cluster shape: identity (what it is, who makes it, license, pricing), decision data (performance, accessibility, vulnerabilities, compatibility), editorial verdict (our call, with the reasoning attached), and activity (release cadence, vendor health, live feeds). The uniform shape is what lets you compare two products on the same axes instead of two marketing pages.
Reproducible Lighthouse scoring
Every Lighthouse score publishes its test configuration alongside the number: the host environment, the simulated network and CPU throttling, the number of runs, and which aggregate we report (median, best, or worst). The default reported figure is the median of repeated runs, not a cherry-picked best.
Raw Lighthouse JSON is available on request for any recorded run. A score without a reproducible configuration is a marketing claim, not a measurement, and we treat it as such.
axe and WCAG conformance
Accessibility scoring runs axe-core against representative templates and records the conformance target it meets (WCAG AA, AAA, or fail) plus the raw violation count. Accessibility is a property of a rendered page, not a product, so we test default and demo configurations and say which one the number came from.
Vulnerability sourcing
Open and historical vulnerability counts come from Patchstack and Wordfence Intelligence and refresh on every build. We surface the open CVE count directly on directory cards and entry pages so an unpatched advisory is impossible to miss. Counts reflect what the upstream databases report at build time; a zero means no known open advisory, not a guarantee of safety.
Tracking whether it is still alive
Release histories track each vendor's published changelog. Vendor activity (blog feed, release atom feed, social presence) is fetched at build time and surfaced on vendor profile pages, so you can see whether a product is actively shipping or quietly abandoned before you commit to it. Abandonment risk is the single most expensive thing to get wrong on a premium purchase, so we make it visible up front.
No pay-to-rank
Vendors cannot pay to be included, ranked higher, or have an entry rewritten. Editorial content carries no affiliate links woven into the copy. The only monetization is a single rotating offer, clearly marked, that sits at the top of select pages and scrolls out of view. Every recommendation in the directory reflects our actual judgement, with the data attached so you can audit the call yourself.
When we get it wrong
Data goes stale and tests have bugs. If a number looks wrong, or a vendor has shipped a fix that our build has not yet picked up, tell us via the contact page and we will recheck and republish. Reproducibility cuts both ways: if you can show the measurement, we can correct it.