Testing electrical products in practice, from a bench sample to a home that uses the product daily

A useful test plan starts with the product as sold, not with a prototype that was hand finished for a showroom. It lists the hazards the team identified, the test that addresses each hazard, the sample that was tested, and the pass condition in plain words. A fan heater plan, for example, should include normal running, blocked vents, a tipped heater, a stalled fan, a damaged cable and a long running session, each with a note saying what safe behaviour looks like. A plan that only lists a laboratory name and a certificate number contains none of that reasoning.

The plan should also name who runs each test, where it is run, and what happens to the sample afterwards. A sample that was tested to destruction cannot then be sold as new, and a sample that was tested only on a bench cannot speak for a product used in a dusty garage. Keep the plan in the file beside the results, so a reader can see not only that a test passed, but what was actually tried and what was never tried.

Bench tests are controlled, repeatable and valuable: they measure insulation, heat, current and mechanical strength under defined conditions, and they let a laboratory compare results fairly. Home tests are messy, realistic and equally valuable: they reveal that a cable is too short for a real socket, that a vent clogs with kitchen grease, or that a control is unreadable in dim light. A product that passes bench tests yet fails home tests is not safe in the sense that matters, because homes are where the product lives.

Samples should come from production, or from a batch that matches production in materials, components and assembly, because a prototype built by engineers often differs from a factory unit in ways that matter. Choose samples across the range sold, including different colours or variants where the enclosure or material differs, and include the charger and accessories as sold, not a laboratory substitute. If only one sample is tested, say so plainly, and explain why that sample represents the range.

Illustration: a seller of electric blankets tested 30 blankets drawn from three production weeks, rather than a single blanket from the first week, and found that blankets from one week had a looser controller cable entry. The blanket count is illustrative, and the lesson is that sampling across production catches drift that a single sample hides. Keep tested samples, or photographs of them, with the test record, so a later question about which version was tested has an answer.

A test report should be read for what it does not say as well as for what it says. Check the product identification, the sample source, the conditions, the version of the product tested, and whether any test was skipped, modified or failed and then passed after a change. A report that lists a failure, a fix and a retest is more trustworthy than a report that shows only passes, because it shows the product was challenged. A report for a different model, a different charger or an earlier version does not prove anything about the product on sale.

A failed test is useful information, not a disaster, provided it is handled openly. Stop, identify the cause rather than the symptom, fix the design or the component, and retest the corrected sample under the same conditions that failed. Record the failure, the fix and the retest together, because that sequence shows the product improved through testing. Hiding a failure, retesting a different sample quietly, or changing the test condition to make the failure disappear all destroy the value of testing and create a file that cannot be trusted.

Where a failure suggests a wider issue, such as a component used across several products, widen the check to those products before deciding the failure is isolated. A single failed heater from a batch may be a one off, but a failed component shared by a heater, a fan and a kettle is a family issue. The file should show that the wider check was considered, even when it concludes the failure was isolated, because that consideration is what separates diligence from luck.

Testing is not a one off event at launch. A new supplier, a new material, a new charger, a new motor or a firmware change can each invalidate part of the test evidence, and the file should be updated accordingly. Keep a short change log that lists each change, the tests it affects, and the retest result. When a change affects only one test, retest that test; when it affects the product broadly, retest broadly. That proportionate approach keeps testing affordable without pretending an old certificate covers a new product.