Testing methodology
How we test, and how to read our evidence.
Every review, benchmark and comparison on this site states how its conclusions were reached. This page describes the method behind them: what gets tested, on what, and what has to be published alongside a result.
Who runs the tests
Kite Hosting is run by Sam Moreton, its founder and editor. The first-hand testing happens on the Cloudways servers he pays for and operates: real production workloads and real bills, not a temporary affiliate test account.
Every page declares its evidence level
Every article, review, comparison, benchmark and tool page carries a visible evidence label near the top, so you can tell what kind of claim you are reading before you read it:
- First-hand testing
- Produced on servers and accounts we pay for and run. The evidence behind each claim is retained and checked before publication.
- Researched
- Built from named sources: official documentation, vendor pricing and reputable third-party material. We have not tested it ourselves.
- First-hand + researched
- A mix of both, and the writing makes clear which is which.
The claim rules behind these labels sit in the editorial policy.
First-hand evidence is retained, not remembered
A first-hand claim is published only when the material behind it (screenshots, invoices, logs, benchmark raw data) is retained in a private evidence vault and has passed our citation gate: owner-approved for publication, with no redaction outstanding. Material that needs redaction becomes citable only through its redacted copy, and private details never reach the published page.
What a benchmark must publish
A benchmark page cannot ship without its test setup, because the content schema rejects one that tries. Every benchmark carries:
| Field | What it tells you |
|---|---|
| Environment | The server, stack and workload that were tested. |
| Methodology version | Which version of our method produced the numbers. |
| Benchmark date | When the runs happened. |
| Runs | How many runs the result summarises. |
| Raw data | Where the raw results are retained. |
| Limitations | What the numbers do not tell you. |
A conclusion without the setup that produced it is not published.
Reviews and comparisons
- Reviews state the verdict near the top, then the period and configuration actually tested.
- Every review states who the product suits and who should look elsewhere, and lists strengths and drawbacks.
- Prices carry the date we last verified them. An unverified price carries no date.
- Comparisons state how far the two products were tested in the same way, and pick a winner per use case rather than a blanket best.
Sources and check dates
Material claims from outside this site name their source, with the publisher and the date we last loaded and checked the page. A check date is only stamped for a URL we actually verified: where a source could not be checked directly, it carries no date rather than a pretended one.
Updates and re-review
Publishing a page records the source snapshot date and a next review date. Pages are re-checked when a product or its pricing changes materially. Dates move only for real editorial review, never automated date bumping, and material changes and corrections are recorded in the page's change log.
What we will not do
- Publish a benchmark number without its environment, method version and run count.
- Fill a gap in the evidence with an estimate: untested is marked as untested.
- Present researched material as first-hand experience.