How we research.

We prefer primary statistical sources — the U.S. Census Bureau, Bureau of Labor Statistics, SBA Office of Advocacy and the Federal Reserve — and peer-reviewed or working-paper research from identifiable authors.

We cite only sources we actually read and used. Every external figure links to where it came from.

Primary evidence is original data or the original study. Secondary evidence summarises someone else's work. When we rely on a summary, we say so, and we go back to the original whenever it is available.

Each paper separates four kinds of statement, and labels them in the text:

External research — published findings from a cited source. Liquid analysis — our own calculation or interpretation of that evidence. Field observation — patterns seen in client and operating work, not statistically representative. Hypothesis — a claim we think is plausible but have not tested.

Opinion is labelled as opinion.

Worked examples and illustrative models state their inputs. They show how a mechanism works, not what any particular business will experience.

Survey definitions change, samples differ and associations are not causes. Where a data series was revised, we note it and do not compare figures across the break as if they were growth. Papers include a limitations note when it matters to the conclusion.

Liquid Research is independently published. It is not peer-reviewed, and papers carry no DOI. We would rather state that plainly than imply an academic process that does not exist.

If a figure, source or claim is wrong, we correct it in the paper and change its last-updated date. Substantive changes — a different number, a withdrawn claim — are noted; typo fixes are not.

Report an error at hello@liquidiq.io.

Liquid Research is published by Liquid IQ, a consulting firm. Findings can relate to services the firm sells. We state that relationship here so readers can weigh it, and papers do not recommend Liquid's services as a conclusion of the evidence.

Where sources are public, each paper's references link directly to the underlying publication or dataset so readers can check the figures themselves.

We use AI tools to help find sources, draft structure and check consistency. A person reads every cited source, verifies every figure against it and is accountable for every claim published.