How We Test WordPress Hosting Speed: Our Methodology

Published October 2026 · 6 min read

Affiliate disclosure: this article currently contains no affiliate links. If we add them later, we will update this notice. Read our full disclosure.

Every hosting review on this site makes claims about speed. This page explains exactly how we arrive at those claims: what we measure, which tools we use, how we run the tests, and — just as important — what we do not claim to measure. If a review ever contradicts this methodology, call it out: the method is the promise.

Our approach in one paragraph: We combine independent third-party benchmark data with hands-on evaluation of each host's platform, and we are building a program of direct load tests on live installs. We publish our methods so you can judge our conclusions. We do not accept payment for test results, and no host sees our findings before publication.

Why Methodology Matters

Hosting marketing runs on speed claims. "20x faster," "blazing performance," "optimized for WordPress" — these phrases appear on nearly every hosting homepage and mean almost nothing, because there is no standard behind them. A host can truthfully claim record speeds while testing an empty page on an enterprise server you will never touch.

Independent testing exists to close that gap. But testing has its own traps: a single speed test on a Tuesday afternoon proves little, and test setups vary so much between reviewers that results are rarely comparable. Our methodology is designed around one principle: measure what a real site owner would actually experience, and be transparent about everything else.

The Metrics We Track

We focus on five metrics, because each one answers a different question a buyer actually has.

1. Time to First Byte (TTFB). How long before the server starts responding. This is the purest measure of hosting infrastructure — it strips out most of what your theme and plugins add. We consider under 200 milliseconds excellent, 200–400 good, and anything consistently over 600 a concern on premium hosting. TTFB is our most-cited metric because it is the hardest for a host to fake and the most comparable across providers.

2. Largest Contentful Paint (LCP). How long until the main content of the page is visible. Google uses this in its Core Web Vitals, which feed into search rankings. Unlike TTFB, LCP reflects the full stack: server, caching, CDN, theme, and images. A host can have great TTFB and mediocre LCP if its caching is poorly configured — which is exactly why we measure both.

3. Fully loaded time. How long until the page finishes loading everything. Less glamorous than LCP, but it captures the tail of slow third-party scripts and unoptimized assets that real visitors experience.

4. Performance under load. What happens when 50 or 500 visitors arrive at once. A host that serves one visitor in 300 milliseconds but collapses at 50 concurrent users is not fast — it is fast in a demo. Load behavior separates genuinely good infrastructure from good marketing.

5. Uptime. The percentage of time the site responds at all. Speed means nothing if the site is down. We track this over months, not days, because a week of perfect uptime proves nothing about a host's reliability.

The Tools We Use

We use industry-standard, publicly documented tools — nothing proprietary, so anyone can reproduce our approach:

WebPageTest for detailed load waterfalls, TTFB, and Core Web Vitals from multiple global locations. Its filmstrip view shows what visitors actually see as the page renders.
GTmetrix for quick performance grades and concrete optimization recommendations, useful for separating hosting performance from theme problems.
Uptime monitoring (independent third-party monitors) pinging test sites at short intervals over months to build real reliability records.
Load testing tools (k6 and similar) to simulate concurrent visitors and find the point where response times degrade.
The host's own tooling — Kinsta's built-in APM, for example — to verify that the performance features marketed on the pricing page actually exist and work.

We deliberately avoid synthetic "hosting speed test" plugins and single-run homepage checkers as primary evidence. They are fine for spot checks; they are not methodology.

Our Testing Process

For each host we review, the process runs in four stages.

Stage 1: Research synthesis. Before touching anything, we collect independent benchmark data published by third parties — reviewers who ran their own tests, with their methods documented. We weight recent data over old data, documented methods over bare claims, and consistent findings across multiple sources over any single spectacular result. This is currently the primary basis for the performance claims in our reviews, and we say so in each article.

Stage 2: Hands-on platform evaluation. We sign up for the actual plans we write about (or evaluate them through documented trial access), install a standard WordPress test site, and work through the real owner experience: the dashboard, staging, backups, migrations, support interactions, and the performance tooling the host provides. A host's speed claims mean more when we have clicked through the panel that supposedly delivers them.

Stage 3: Direct load testing (in progress). We are building dedicated test installs on each major host to run our own controlled measurements: identical WordPress setups, same theme and content, tested from the same locations, over weeks rather than hours. Where our reviews currently say "independent benchmarks report…," this program will eventually let us say "we measured…." Until those numbers exist, you will see placeholder notes in our reviews marking exactly which claims await our own data. We would rather show the gap than hide it.

Stage 4: Long-term monitoring. Speed at signup is not speed at month six. We keep test sites running and monitor uptime and response times over time, because hosting quality degrades more often than it improves — oversold servers, neglected infrastructure, support teams that shrink after acquisition. A review is a snapshot; monitoring is the check on whether the snapshot stayed true.

How We Keep Tests Fair

A few rules we hold ourselves to:

Same test site everywhere. Our direct tests use identical WordPress installations — same version, same theme, same plugins, same content — so differences in results reflect the hosting, not the setup.
Same locations. We test from multiple geographic regions and report where, because a host with one US data center will look amazing from New York and mediocre from Sydney.
No cherry-picking. We report sustained and average figures, not the single best run. If a host is fast nine times and slow once, you hear about the slow one too.
Caching on, as shipped. We test with the host's default caching configuration, because that is what a real customer gets. We note when we change any default.
No paid placements. Hosts cannot pay for better results, preview our findings, or see drafts. Our revenue — when affiliate relationships exist — comes from reader signups, never from review scores.

What We Do Not Measure (and Why)

Honesty requires boundaries. A few things we deliberately do not claim:

We do not test every plan. We test representative tiers — usually entry and mid-range — and say which. Performance on a $1,650 enterprise plan is a different product.
We do not measure your specific site. Our test site is standardized; your WooCommerce store with 40 plugins will behave differently. Our numbers tell you about the platform's capability, not your exact outcome.
We do not review support by one chat. Support quality is assessed across multiple interactions and corroborated with independent customer feedback, not a single conversation.
We do not re-test constantly. Hosting changes slowly. We revisit published reviews when hosts change plans, infrastructure, or ownership — not on a fixed schedule.

How to Read Our Reviews

With this methodology in mind, here is how to get the most out of our articles:

Check the date. A 2024 benchmark cited in a 2026 review is context, not news. We date our claims.
Look for the placeholders. Where we write "[independent benchmarks report…]" versus "we measured…," the difference tells you exactly how much of the claim is ours.
Weight TTFB for infrastructure, LCP for experience. If you are choosing between two fast hosts, LCP and load behavior matter more than a 30-millisecond TTFB gap.
Read the caveats. Every one of our reviews includes an honest-caveats section. The most useful sentence in a review is often the one that starts with "but."

The Bottom Line

Good hosting testing is slow, boring, and transparent — the opposite of hosting marketing. Our methodology is built to be checkable: standard tools, documented setups, published limitations, and clear labels on what we measured ourselves versus what independent sources report. As our direct testing program grows, more claims will move from "benchmarks report" to "we measured," and we will update published reviews when they do. Until then, we would rather be precisely honest about what we know than impressively vague about what we do not.