Three inputs, weighted roughly equally: what readers ask for by email, what we see failing in client work, and what is genuinely new rather than merely newly announced. We deliberately ignore launch calendars — a component that has been on sale for eight months usually has more useful data behind it than one that shipped last Tuesday.
If you want something covered, tell us. Reader requests drive about a third of what we publish.
Sometimes. When they do, we say so in the article. Review samples are returned or, when the manufacturer will not take them back, donated to a local repair café and logged publicly. We buy the majority of what we test at retail, which is slower and more expensive but keeps the relationship simple.
Usually because the test conditions differ. Ambient temperature, case airflow, driver version, power plan, background services and even the order in which tests are run all move the numbers. We publish our conditions with every article precisely so you can see where a difference comes from. If our numbers look wrong to you, email the specifics and we will re-run the test.
That is the intention. Every benchmark article lists the exact tool versions, settings and run counts. Where a test uses a script we wrote, the script is published alongside the article. If a result cannot be reproduced from what we published, treat that as a bug and report it.
No. Sponsors receive a publication date and a fact-check window limited to factual corrections about their own product — specifications, pricing, availability. They cannot see or influence the verdict, and we do not remove or rewrite live posts. This is written into every agreement.
With a visible label at the top of the article, a label in every feed and listing where it appears, and rel="sponsored" on outbound links to the sponsor. The label never disappears over time. If you cannot tell within two seconds that something is sponsored, we have failed.
Around 210,000 monthly page views and roughly 9,400 newsletter subscribers, with a 41% average open rate. About 61% of traffic is North America and Western Europe. We will send the full breakdown, including the unflattering parts, to anyone considering a placement.
No, none of the three. Ad creative is served from our own origin, so no third-party scripts run on the page. This limits what we can charge and rules out some advertisers entirely, which we have made peace with.
Up to ninety minutes of our time at no cost. We look at the machine, site or account, identify what is actually wrong, and send you a written summary with an estimate. You are free to take that summary to another provider — several people have, and we would rather that than sell you the wrong thing.
Yes. Roughly 40% of our client work is asynchronous. We keep one overlapping hour per working day for calls and handle everything else in writing, which most clients end up preferring because it leaves a record.
Tell us within thirty days and we will redo the work or refund the fee — your choice. We have honoured that four times in ten years. The condition is that the original scope has not changed; we cannot refund a project because the goalposts moved.
You do, entirely, from the moment the final invoice is paid. No licence-back clauses, no dependency on tooling we control. We will hand over repositories, credentials and documentation in whatever form you want them.
Anonymous, aggregated page-view counts and, if you submit a form or subscribe, only the details you typed. There is no advertising cookie, no cross-site identifier and no session replay. The Privacy Policy spells out every category, and the Cookie Policy lists each cookie individually.
Not in full, but you are welcome to quote up to 200 words with a link back, and to use our charts in presentations with attribution. For anything larger, ask — we usually say yes for educational and non-commercial use. Details are in the Terms of Use and the DMCA page.
Because there is almost nothing on it: self-hosted fonts subset to what we actually use, one stylesheet, one small script, no framework, no third-party tags. It is also the honest answer to a fair question — we write about web performance, so the site had better be quick.
Email [email protected] or use the contact form and pick “Report a correction”. Substantive corrections get a dated note at the foot of the article. We do not quietly edit numbers.
Still stuck?
If your question is not here, send it over. We answer everything within one working day, and genuinely useful questions tend to end up on this page.
Ask us directlyGet the Monday Byte
Every Monday morning: one deep-dive guide, one tool that earned its place on our bench, and one benchmark number that surprised us. Around 9,400 readers, zero fluff.
By subscribing you agree to our Privacy Policy. One click unsubscribes you for good.