How we work

Two kinds of page live on this site: reports built from primary statistics, and guides that rank software. This is the method both are built to. If a page does not follow it, the page is wrong, not the method.

We start at the measurement, not at the article about the measurement. A report begins with a named dataset from a statistics office, regulator or supervisory body, and a script in this site’s repository queries that dataset directly. The script writes the page. That is the whole chain, and it is short on purpose:

• Primary source or nothing. If the only trace of a number is a blog post citing a press release, it does not go on the page.
• The query is the method. Because the figures are fetched rather than typed, a reader who opens the dataset with the same filters lands on our number.
• The period travels with the figure. Every data point states its reference period, and every source states when it was last published by its publisher.
• No silent rounding into confidence. Percentages are shown as the source reports them, to one decimal where the source has one.
• Derived figures are marked as derived. A ratio or a gap we calculated from two published figures is labelled, not presented as a measurement.

Every data point carries one of three:

• Verified. Read from the named primary source.
• Directional. The sources agree on the direction but measure it differently, so the number is indicative rather than exact.
• Single source. One publisher reported it and nobody has corroborated it. We would rather say that than let a lone figure pass as settled.

Data points are numbered, and the numbers are addressable: data point seven on a report has its own anchor and keeps it. Every report carries ready-made APA, MLA and Chicago citations, and the same figures appear once more in the market data index with their source and period. If a figure is corrected, the number stays and the correction is noted, because a citation that quietly changes underneath the person who made it is worse than no citation.

Every page carries a writer, an editor and a fact-checker, named at the top. The writer builds the page. The editor decides whether it is finished. The fact-checker takes every specific claim — a figure, a price, a country, a certification — back to the source it came from. Nobody edits or checks their own writing, and which role falls to whom is fixed per page, so a page keeps the same three names when it is rebuilt.

Feature checklists are the least useful way to compare software, because every vendor ticks every box. We rank on what separates products after the demo:

• Time to useful. What the first ninety days actually involve, and whether you need an implementation partner to get there.
• The real price. What the per-seat number becomes once the modules a normal buyer needs are added, and what happens at renewal.
• The exit. How your data comes out, in what format, and whether that export is included or billed as a project.
• Who it is for. A product that is wrong for you is not a bad product, and we say which is which rather than declaring a single winner.

Vendor claims are read from the vendor: the imprint, the privacy statement, the pricing page, the trust centre. Country means where a vendor is headquartered or contracts from, which is a different question from where your data is hosted; where those two tell different stories, the entry says so. Where a company will not publish something — its legal entity, its price — we write that on the page rather than filling the gap with an assumption.

Positions one to six in a ranked list are commercially available. When a vendor holds one, it changes the order in which products appear. It does not change whether a product is listed, and it does not change one word of what the entry says, including the criticism. Every ranked list carries that disclosure above it, not in a footer, and the full arrangement is on the services and disclosure pages. Research reports are never sold this way: a commissioned report names its commissioner on the page.

If something here is wrong, tell us and we will fix it and say that we did. Errors of fact get corrected on the page; errors of judgement get argued out in the text rather than quietly rewritten. It makes no difference whether you are a reader, a competitor or the vendor being criticised. Write to business@gcas-digital.com.

Each page shows when it was published and when it was last reviewed. For a report, a review means the source dataset was queried again and the figures rewritten if the publisher revised them. For a guide, it means someone went back to the vendor documentation. In neither case does it mean a date was bumped.