How we test and rank
Updated
Every product in a ranking or comparison is judged the same way, whoever makes it. Where a product is made by the company that publishes this site, the page carries a short note in the fine print, and the product is scored with the same rubric as everything else.
Desk reviews and hands-on tests
Some of our guides are desk reviews: we compare what each maker publishes about its own product (pricing pages, help centres, feature lists) and link every one of those pages. A desk review says so at the top, gives no scores, and stays out of search results until it has been tested.
A hands-on test names the tester and states how long they used each product, on what hardware and in which DAW. Only hands-on tests carry scores.
The scoring rubric
Every hands-on test scores each product out of 100 on the same five criteria. The weights are fixed before testing starts and are the same for every product in a list. A best-of page names the criteria it uses in these terms.
| Criterion | Weight | What it covers |
|---|---|---|
| Performance | 35 | How well the product does its core job, measured by us (listening tests, null tests, measured loudness, latency, stability). |
| Rights and data | 20 | Who owns the output, what the terms let the maker do with your files, where they are processed and how long they are kept. |
| Cost and limits | 20 | What a realistic month or release costs at published prices, and how far a free tier really goes. |
| Workflow and formats | 15 | Input and output formats, DAW integration, collaboration, batch work and speed. |
| Control | 10 | What you can change after the first pass, and whether the changes do what they say. |
Evidence. A score point needs evidence we can point to. Performance comes from our own tests, with the method written on the page. Rights, cost and workflow come from the maker's published terms, pricing and help pages, each linked and dated on the page. A maker's own claim is recorded as a claim until we have tested it. We score what we found, not what the marketing promises.
Unknowns. If we cannot find or test something, the page says "not found" or "not tested", never a guess, and that criterion gets no points for the missing part. A product is not penalised below what we could verify, and we do not fill gaps with an average.
Criteria that don't apply. If a criterion does not fit a category (for example, Control for a tool with no settings), the page says so and the other weights are scaled up in proportion. The page shows the weights actually used.
Ties. Products within two points of each other are shown as tied, in alphabetical order, with the individual criterion scores beside them so readers can weigh what matters to them.
Changes. The rubric is versioned. When we change a weight or a criterion we say so in the corrections log and re-score the affected lists, rather than quietly shifting scores. Fixed weights per version follow the approach of RTINGS, which weights each test by its relevance to a usage and makes every test and weight change at once, as a new versioned test bench with a published changelog.
What never affects a ranking
Payment never changes where a tool appears or how it scores, and payment data is never an input to a score, a ranking or a best-of list. Paid directory placements are labelled "Featured" or "Sponsored", and paid and affiliate links are marked rel="sponsored".
Corrections
When we get something wrong we fix it on the page and list it in the public corrections log.