An independent testing lab
triedandtested.org is an independent lab for testing digital services. We are not sponsored and do not participate in any affiliate programs.
No link on this site earns us an affiliate commission. Our comparisons are based on our tests and the sources cited in each report.
01 — THE METHOD
What we test, and what we do not
We rent the service as an ordinary customer: at the public price, without telling the provider, with no courtesy access and no press account. A provider who knows it is being watched does not behave as usual.
We test what a customer can establish for themselves: what is delivered, what is billed, and what happens when something goes wrong. That is the boundary of what we can prove.
- Out of scope
- A provider’s internal security, its financial health and the sincerity of its intentions. None of that can be observed from a customer account, so we do not comment on it.
02 — THE METHOD
The conditions behind each result
Each report states the test duration and the conditions available in the lab record: plan, configuration, country, price and payment method where recorded. Catalogue prices are identified separately from amounts actually paid.
The period covered is stated service by service. A service followed for three weeks and one followed for six months are not comparable, and we do not put them in the same column without saying so.
When a plan changes mid-test, the version tested is the one that counts, and the change is dated. We do not apply new conditions retroactively to old facts.
03 — THE METHOD
What we keep
The reports draw on the lab’s test records. Their tables and downloadable datasets collect the recorded outcomes; links to provider documentation supply the published offer details.
The publication distinguishes measured counts, qualitative observations and overall lab ratings. An unavailable value remains unrecorded, and no score is inferred from it.
- What is not published
- Account credentials and personal data are excluded from the public records. The downloadable datasets contain the test results and the source details recorded for publication.
04 — THE METHOD
Three levels of reading, never mixed
This distinction is the backbone of the site. Every published data point carries its level, and one level is never converted into another for the convenience of a sentence.
- Observed during the test
- A result or event recorded by the lab during a test. Its scope follows the duration, counts and conditions documented in the report.
- Stated by the provider
- Taken from its documentation, with the page address and the date it was consulted. A written commitment describes a published intention, not observed behaviour.
- Not verified
- Not tested, or the record is unavailable. The cell stays empty and no conclusion is drawn. A missing measurement is not a bad result.
05 — THE METHOD
How a ranking is built
Each report states what determines its order: a measured result or an overall lab rating. If a weighted formula is used, its calculation is published.
Missing data stays unrated. It does not become a zero: a service we did not measure is not a bad service, and pushing it down the table would be a methodological error.
An overall rating expresses our editorial assessment of the experience. It is identified as such and accompanies the test results; it does not become a measurement of reliability.
06 — THE METHOD
What invalidates a test
Some events strip a test of its comparative value. We would rather say so and start again than publish a result we know to be fragile.
- The provider learns it is being tested
- If it changes its handling from that point, the observed period is no longer representative. The test restarts from scratch on a new account.
- One server does not represent a plan
- Behaviour observed in one country does not carry over to the same provider’s other locations. We name the country tested and confine ourselves to it.
- Conditions change during the period
- The result is tied to the version of the terms in force when the events occurred, with its date. Both versions are cited when the difference matters.
07 — THE METHOD
Corrections and provider responses
Every correction is dated and flagged on the entry concerned. We do not rewrite silently: what was corrected stays legible afterwards.
A provider that contradicts a record is quoted, with its response and its date. If that response establishes an error on our part, the entry is corrected and the correction is visible.
A takedown request that brings no factual element does not make a record disappear. In that case we publish the request and our answer.
08 — THE METHOD
What a test does not prove
Behaviour observed over six months guarantees nothing about the next six. A host changes policy, ownership, transit provider, jurisdiction.
The absence of an event during a test does not prove it impossible. Having received no request for identity documents does not mean there never will be one.
Our results describe what happened, where and when. They are not legal advice, not a guarantee, and not a recommendation to use a service in any particular way.