Data

How Indexed.vc data is built

What we accept as evidence, what we refuse to guess, and how rankings and reports are assembled.

Sources

Every fact on a company or investor profile traces to a page we fetched: the company's own website, its LinkedIn organization page, a press release or wire announcement, a regulatory filing (for example SEC Form D or SBIR award records), or an investor's published portfolio. Search-result snippets, third-party summaries, and model memory are not sources. Where a profile cites a press announcement, the link is shown on the round.

What is never inferred

  • Funding amounts are recorded only when a source states them. An unknown amount stays blank; it is never estimated from a headline or a valuation.
  • Amounts are stored in whole US dollars. A round announced in another currency keeps its native amount, and is converted only when a source states the USD figure.
  • Founding years, headquarters, headcount bands, and industries come from a page that states them. An accelerator batch date is not a founding date.
  • Acquisitions, IPOs and shutdowns require a source naming the event. A parked domain or a stale website is not evidence of closure.
  • Technology and tool usage is recorded only from first-party evidence such as job postings, engineering pages, or attributed case studies — never from what a product integrates with.

Verification before publication

Research passes through deterministic gates before anything is written: field validators (URLs, enums, taxonomy, dates, currency), a grounding check that every number, month, and named claim in a description appears verbatim in the quoted source text, and a provenance check that re-fetches each cited page and confirms the quote is on it. A sample of every batch is then independently re-read against its sources; a batch whose sample fails is reviewed in full, and any record with an unsupported claim is withheld rather than published with the claim removed silently.

Identity is resolved on durable anchors — canonical domains and LinkedIn organization identifiers — before names. Two companies that share a name are never merged on the name, and a round is never attached to a company on the name alone.

Rankings and reports

Each ranking states its filter and its ordering rule on the page: which sectors, countries, or funding stages it includes, and whether rows are ordered by disclosed total funding or by most recent announced round. Only companies whose funding is disclosed can rank by funding. Every row links to the profile that cites its source. The date shown on a list is the newest change to any company on it, not the time the page was rendered.

Reports are aggregates over the same records — totals, stage mixes, sector and geography cuts, investor league tables — computed from rounds that carry a stated amount and a date. They are recomputed continuously as rounds are added or corrected.

Coverage and freshness

Coverage grows through observed demand — what people search for, look up, and ask us to add — and through discovery from investor portfolios, filings, and announcements. Profiles carry the date of their last verified update. A profile that has not been verified recently says so rather than presenting old facts as current, and a company we cannot resolve to a durable identity is held out of the public index until we can.

Corrections

Anyone can submit a company or report a correction. Corrections are applied through the same evidence gates as new research: a change needs a source stronger than the one already cited.