Two joinable datasets on every unicorn startup Wikipedia lists: all 836 companies (628 current unicorns and 208 former ones, with their exit date, reason and exit valuation) and 1,123 founder rows, one per founder per company exactly as the list credits them. Companies carry the list facts (valuation in US$ billions with its date, industry, country) plus everything their own article adds: company type, founded year, headquarters split to a country, products, revenue, net income and employees with numeric twins and years, website, and the FUNDING story mined from each page, funding-round wikitables where they exist, else the Funding section's prose: rounds counted, first and last funding years, total raised in US$ millions preferring an article-stated total, and lead investors de-duplicated in order of first appearance. Derived: Years_To_Unicorn (median 8), Founder_Count and Valuation_Per_Employee_Millions. Founders resolve their articles through three tiers, the list's own link, the company infobox's founder link, then a name-guess kept ONLY if the article mentions the founder's company, so a same-name stranger can never wear a founder's biography; the 242 resolved get birth and death dates, age at founding, birthplace split per house style, marriage and children counts, gender by pronoun analysis, citizenship, occupation, net worth where stated, and an education read: alma mater, graduate school, highest level classified from degree tokens, and a dropped-out flag from the education cell or the article's own words. Unresolved founders keep their rows with honest NAs; nothing is guessed from a name. Companies_In_List flags the serial founders. Built with a 16-test acceptance suite covering join integrity, sanity ranges and known-fact anchors.
1,959 rows across 2 joinable files, with 52 columns in total: Company, Status, Valuation_USD_Billions, Valuation_Date, Industry, Country, Founders, Founder_Count, Company_Type, Founded_Year, Years_To_Unicorn, Headquarters, HQ_Country, Products, Revenue_USD_Millions, Revenue_Year, Net_Income_USD_Millions, Net_Income_Year, Employees, Employees_Year, Valuation_Per_Employee_Millions, Website, Funding_Rounds, First_Funding_Year, Last_Funding_Year, Total_Funding_USD_Millions, Lead_Investors, Exit_Date, Exit_Reason, Exit_Valuation_USD_Billions, Company_URL, Founder, Company, Gender, Birth_Date, Death_Date, Company_Founded_Year, Age_at_Founding, Birth_Place, Birth_Country....
It is built by programmatically scraping en.wikipedia.org, last pulled on 2026-09-11. Datasets are versioned; older versions stay downloadable.
Yes. Every CodeSights dataset is completely free as a CSV download; a free account is all it takes. Anyone can preview the data without signing in.
Yes. The exact Python scraper that built it is viewable on the dataset page by any signed-in member, so every number is reproducible.
Join the files on company name (founders.Company = companies.Company); Founder_Count on the companies file says how many founder rows to expect.
Automated scraping leaves room for error and the underlying sources change over time, so no version is guaranteed accurate or complete. If a number matters, verify it against the original source.