How the ranking is built
We run no tests of our own. We read the ones that exist, whatever language they were published in, and merge their results into one score per product.
One common scale
Every source scores differently: school grades in Germany, percentages in the UK and the US, marks out of 20 in France. We convert each result to 0–100 points, higher is better. The original verdict is stored unchanged and shown next to every source.
Overall verdict and criteria
A product’s overall score is the weighted mean of the overall verdicts of all tests we reviewed. Where a test also publishes individual results, we map them to the category’s criteria, such as cleaning performance or battery life. A criterion’s value is the weighted mean of all tests that assessed it. If no test assessed a criterion, it is missing for that product and is not counted as zero. If a test gives only individual results, we derive its overall verdict from those – but only if they cover at least two criteria and more than half of the weighting. A single partial result counts only towards its criterion; the table of tests then shows no converted value.
Sources carry different weight
Lab tests from consumer organisations count in full, editorial tests without a published protocol with a factor of 0.6, meta-reviews with 0.4 and customer ratings with 0.25.
Your weighting
The sliders change how much each criterion counts. The score then shifts by the difference between your weighted criteria mean and that of the editorial weighting. A device can only follow your own weighting if tests provide individual results for at least half of it. Otherwise it appears below the ranking as “not ranked” – with its overall score, but without a position. If a smaller part is missing, it stays in the ranking and the missing criteria are listed. We do not estimate missing values; each slider shows for how many devices test values exist. The calculation runs in your browser with the same data and the same formula.
One score, several markets
A product’s score is identical in every language version. What differs is price and retailer. A model with no offer on file for a market stays visible, without a buy button.
Contradictions stay visible
When tests reach different conclusions, we do not smooth that over. The spread chart shows the average result per test country, and every product lists each source with its country and original verdict.
Test basis
Three small bars next to the score show how broadly a device has been tested: “broadly tested” from six tests in at least three countries, “thinly tested” with fewer than three tests or tests from only one country, and “solidly tested” in between. The test basis does not change the score; it tells you how much stands behind it. It is different from the individual results per criterion, which decide whether a device can follow your own weighting.
Notices
Alongside the tests we show notices for individual devices or brands – such as allegations, recalls or commitments made by a manufacturer. Every notice names its sources, the date, its current status and, where available, the manufacturer’s response. An allegation is labelled as an allegation; we do not verify it ourselves. We only include notable cases – safety, privacy, serial defects, lawsuits and regulatory proceedings, and binding commitments – and only where an official source or several independent reports exist. The absence of a notice does not mean a device is free of problems. Notices have no influence on score or order.
Affiliate links
Shop links are affiliate links. Commissions affect neither score nor order; both result solely from the calculation described here.
Converting German school grades
Values between the anchor points are interpolated linearly. Percentages are used as they are; ratings out of 5, 10 or 20 are converted proportionally.
| Grade | Points |
|---|---|
| 0.5 | 100 |
| 1.5 | 90 |
| 2.5 | 75 |
| 3.5 | 60 |
| 4.5 | 45 |
| 5.5 | 20 |