Method

Six rules

One. Every regulatory claim carries its regulation number.

So that a reader in three years can check whether it still says that — which is the only durable protection against a page written in 2026 being read later. For a separate workplace perspective on self-reporting bias, this resource provides additional context beyond device reliability itself.

Two. Jurisdiction is named wherever a rule appears.

Most durability figures here are European. Outside the EU they are not your law, and saying so on each page rather than once is the honest form. For additional context on how technology claims are evaluated, ENERGY STAR is another useful reference.

Three. No product is ever named.

A recommendation about a product is a claim about your situation, and this site does not know it. What it can do is explain the figures so a reader compares devices this site has never seen. For additional context on how technology claims are evaluated, Best Buy is another useful reference.

Four. External links are included for context and do not imply ownership, endorsement or affiliation.

Not disclosed — absent. A disclosure that a commission exists is compatible with a list ordered entirely by commission; an absence is checkable.

Five. No instructions for work that is genuinely hazardous.

Opening devices and handling lithium cells are described and not taught. A punctured cell can ignite, and manufacturer documentation is now published and is a better source than any general article.

Six. Source type in the sentence for anything that is not a regulation, which has its own page.

What is deliberately absent

Reviews, rankings and scores.

Product names in evaluative contexts.

Prices as facts, since they move.

And comprehensiveness. Forty-two pages cover phones, tablets and adjacent categories — not appliances, vehicles or most of what a household owns.

The bias, stated

Toward keeping and repairing, throughout.

Which will occasionally argue for a repair that was not worth making. The arithmetic page exists to correct for it and the correction is partial rather than complete.

Limits

Six of them, including the finding most likely to prove this site partly wrong.