Methodology published · Sites not yet provisioned · Results: none published
What goes on every host
One WordPress installation per provider, on the cheapest shared or entry plan sold to the public at the time of purchase, bought at retail with our own money. Every site runs the same release, the same default theme, the same twelve posts and images, and no plugins beyond what the host installs by default. No CDN is added and no host-specific caching is enabled beyond what the plan ships with — the point is the plan as sold, not the plan as tuned.
Each site lives on its own subdomain of a domain we control, on the host's default nameservers, in the host's default region for a US buyer where a choice is offered. Plans, purchase dates, renewal dates, and monthly cost are recorded in the study ledger and published with the results.
What we measure
- Availability and TTFB, continuously. FourSight probes in the United States, Canada, Europe, and Asia-Pacific request each site's home page every five minutes with the same quorum rules as the uptime study: a site is down only when a majority of regions agree, and blocked responses are treated as unmeasurable rather than down.
- Browser performance, weekly. A Lighthouse run from a residential connection on our own hardware, desktop preset, three passes, median reported: performance score, server response time, first and largest contentful paint, total blocking time, layout shift, and transferred bytes.
- Consistency over time. The spread of a host's weekly numbers matters as much as the average. Shared hosting's real failure mode is the noisy neighbour, and only a time series shows it.
What we deliberately do not do
- No load testing. Shared-hosting terms forbid it and it would degrade service for the other customers on the box. We measure what a normal visitor experiences.
- No tuning. We do not open support tickets asking for optimisation, enable paid caching tiers, or install performance plugins. If a host's default setup is slow, that is the finding.
- No anonymity games. Sites are ordinary customer accounts. If a host identifies and treats a study site differently, that becomes part of the record.
Publication and exclusions
The roster freezes when the last site is provisioned and is published on this page with purchase dates. Results follow a 30-day baseline, then quarterly. A host whose plan lapses, suspends the account, or blocks our probes is reported as such with dates, never silently dropped. Measurement artifacts — probe outages, our own configuration errors — are excluded and disclosed here, as in the uptime study. Any change to this methodology is dated below.
Budget disclosure: at roughly $3 to $12 a month per plan, the study costs $400 to $700 a year in hosting. It is funded from the site's affiliate revenue, and affiliate status has no bearing on which hosts are included — the roster is every shared host we cover.
Schedule
- September 25, 2026
- Methodology published (this page).
- October 2026
- Sites provisioned; roster and purchase ledger published here.
- 30 days after roster freeze
- First baseline results.