Research methodology

This page explains how we research, test and analyze. Each research article or review also describes its own method.

Sources

We prefer primary sources: laws and official texts, regulators, central banks, court decisions, academic papers, company filings, technical documentation and original datasets. Secondary sources (media, third-party reports) are used to find leads or for context, and are identified as such.

For each source we record the publisher, the link, the publication date (when available) and the date we accessed it. When a rule depends on the country, we say which jurisdiction it applies to.

Data and numbers

• We say where every number comes from and the period it refers to.
• Estimates and projections are labeled as estimates, with who made them and on what basis.
• Market data (prices, volumes, rates) always comes with the source and the date in the caption.
• We do not round, cherry-pick periods or change scales to make a result look bigger or smaller than it is.

Testing tools and products

When we review a tool, we record the version tested, the dates, the environment, the tasks performed and how access was obtained (paid by us, free plan, trial or provided by the vendor). Pricing is checked on a stated date. Vendors do not see the review before publication.

Research projects

Original research states its question, data sources, collection period, method and known limitations before presenting results. When possible, we publish the data or explain why it cannot be published.

Uncertainty and limits

We say what we do not know. When evidence is weak, contradictory or recent, the article says so. Fact, estimate, hypothesis and opinion are kept apart.

AI in our research

AI tools can help us search, summarize documents or write analysis code. Every data point that reaches an article is checked against the original source, which is listed in the article itself. AI output is never cited as a source.