About
Statistics you can trust, from data you can cite
Findings turns authoritative public data into real statistical findings, explained in plain language, with every number traceable back to a computed result. It exists for the moments when a plausible-sounding answer isn't good enough and you need one you can defend.
What makes it different
The hard, risky part of data analysis isn't writing prose; it's getting trustworthy data, computing rigorous statistics, and being able to show your work. Findings is built for exactly that. Every finding is a real test (a correlation, a group comparison, a trend, a chi-square, or an ML pattern), reported with its sample size, p-value, effect size, and the exact query that produced it. The AI summarizes and answers questions about those results, but it is blocked from inventing numbers.
Built to be read by anyone
Results lead with a plain-language summary and the key patterns, then let you ask follow-up questions in a grounded chat. The full methodology (fields analyzed, data dictionary, and test details) is always one click away, so technical and non-technical readers get what they each need from the same report.
Analysts & researchers
Defensible numbers fast, without building the data pipeline yourself.
Journalists & policy
Findings you can cite, each tied to a computed result and its source dataset.
Everyone else
PMs, comms, and domain experts get plain-language answers and a grounded chat, with no SQL required.
Where the data comes from
Findings connects to authoritative public sources: the U.S. open-data catalog (data.gov), the Federal Reserve's economic data (FRED), and the World Bank. Each dataset carries its license and attribution, and only ingestible, well-formed files are offered for analysis.
Try it on a question you care about
Pick a public dataset or two and get ranked, verifiable findings in a couple of minutes.