Sources & Methodology

What we prefer, and why

We weight sources roughly by how directly they connect to the underlying fact. A government dataset or an official regulator statement is preferred over a news article summarizing it. A peer-reviewed study is preferred over a blog post citing "research shows." A named expert speaking on the record is preferred over an anonymous forum post, even a widely-shared one.

Full preference order: government publications and regulator material; peer-reviewed research; universities and academic institutions; official company documentation; recognised professional organisations; original datasets; reputable specialist publications; experienced named experts; high-quality journalism; and clearly identified first-hand accounts.

What we avoid relying on

We avoid building an article's core claims primarily on anonymous blogs, content farms, AI-generated sources, unsourced social media posts, low-quality affiliate sites, quote aggregators, search-result snippets taken out of context, or other articles that themselves lack primary sourcing. Public discussions (forums, comment sections, social platforms) may inspire a question worth answering, but we don't republish them as reporting.

Handling time-sensitive information

For anything that changes over time — a regulation, an economic figure, a workplace statistic — we try to record the source's original publication date, the date we retrieved it, the jurisdiction it applies to, and any known limitations in the data (sample size, methodology, date range). This is especially important in our World, Decoded section.

Labelling first-hand experience separately

Content based on the founder's own business or decision-making experience — mainly our Field Notes — is labelled explicitly as personal experience and analysis, not as generalised, evidence-backed fact. We don't dress up one person's experience as a universal finding, and we don't cite it the way we'd cite a study.

Research limitations

Where the underlying evidence for a topic is thin, mixed, or contested, we say so directly in the article rather than picking the most confident-sounding source and presenting it as consensus.