Crunchbase

Crunchbase is the record of who funded what, who bought whom, and who just took which job - held as profiles for millions of companies, with the operator declining to publish a verified count. Five families of records hang off each profile: funding rounds, acquisitions, investors and the people behind them, subsidiaries and portfolios, and the growth-signal stream of leadership hires and product launches that fills the gaps between deals. On top of the records sits the part its homepage leads with: acquisition-prediction, funding-likelihood and growth-insight scores computed over the corpus.

Inside industrial conglomerates research it plays a role nothing else in the slice plays. The filings layer - SEC EDGAR above all - describes listed parents. The macro layer - BEA, Eurostat, World Bank - describes industries. This is the only one of the 23 primary industrial-conglomerates datasets that covers private-market financing: the scale-up raising a Series C, the grid-equipment maker absorbed by a rival, the investor network connecting them. Coverage is global with its deepest concentration in US and European venture-backed companies. See where it sits on our industrial-conglomerates data hub.

Illustrative rows in the delivered shape - one record family per row:

# one row per record - company, funding round, investor, person, acquisition or signal

record_type : funding round      company       : <industrial-robotics scale-up>
stage       : Series C           amount_raised : on request
lead_investor: on request        event_date    : on request

record_type : acquisition event  company       : <grid-equipment maker>
counterparty: on request         direction     : acquired-by

record_type : person             company       : <robotics scale-up>
full_name   : on request         role          : leadership hire

record_type : company            hq_location   : on request
sector_tags : [industrial automation, energy equipment]
signal      : acquisition / funding-likelihood / growth scores on qualifying profiles

Read the anatomy rather than the placeholders. A Series C row, an acquisition row, a person row and a bare company profile all share one spine - a company key that every family resolves back to. That is the property that turns three decades of scattered events into a panel: stack the rows and any given company's funding history, M&A exposure and leadership churn assemble themselves in order, no entity-resolution pass required. Live values for the companies you name arrive with the sample.

Ten fields span the five record families, keyed on the company spine: identity (company_name, company_key), placement (headquarters_location, sector_tags), the dated events (event_date, amount_raised, counterparty_name, role_or_stage) and the predictive layer (signal_score). One row carries one family, so a funding round and the leadership hire six months later read as two rows of the same company rather than two different databases.

Provenance, stated plainly: the operator publishes no field-level data dictionary, so these definitions trace to the platform's observable surfaces - profile pages, search filters, event feeds - rather than to a published schema. They hold across the corpus, but the exact column names and typed extract schema fold under additional fields on request and are pinned down against a delivered sample before anything depends on them.

Geography - global by design, unevenly deep by gravity: the strongest concentration sits in US and European venture-backed companies, which is precisely where industrial-tech scale-ups live. A Munich robotics Series C and an Austin sensor round resolve equally well.

Temporal - funding events reach from the late 1990s to the present, roughly three decades of private-market activity. That span crosses the dot-com cycle, the 2010s industrial-software wave and the recent climate-hardware run inside one consistent record, which is what makes longitudinal questions answerable at all.

Granularity - five record families descending from company profile to individual event: per company, per funding round, per investor, per person, per acquisition event. Deals and leadership changes stay individually addressable instead of dissolving into annual aggregates - the difference between reading a trend and dating one.

API, files, or your warehouse. Daily, weekly, or hourly.

Cadence is your call even though deal events land when they land - load the historical depth once as a backfill, then keep new rounds, acquisitions and hires rotating into place on whatever schedule your models and alerts expect. Rows arrive flattened to the ten-field shape above with company identities already normalized, so joins against your CRM or feature store need no fuzzy-matching pass. A sample cut to your named companies comes first either way.

• Competitive early-warning - a rival's fresh round, new board seat or quiet acquisition surfaces as a row long before it becomes a press-cycle narrative; see competitive intel product teams use cases.
• Private-target screening - investors trace funding rounds, acquirers and investor networks across private industrial-tech names back to the late 1990s; see investors quants use cases.
• Outbound prioritization - sales teams rank accounts by recency of funding and leadership movement, because a company that just raised buys differently than one that just restructured; see sales growth teams use cases.
• Landscape sizing - researchers count emerging players per sector from funding and acquisition records instead of stitching trade-press anecdotes; see market researchers use cases.
• Feature enrichment - data scientists join funding and investor history onto firm records as model features; see data scientists use cases.

Three personas hold this dataset at relevance 3 of 3 in their industrial-conglomerates packs - sales and growth teams, investors and quant researchers, and competitive-intelligence teams - an unusually strong showing in a slice where most records serve macro analysts. The reason is structural: everything else in the vertical watches companies that already filed; this one watches the ones that have not yet.

Market researchers and consultants (relevance 2) get sector landscape maps with player counts they can defend. Data scientists (2) get event-level funding history as features rather than as prose. Developers and builders (2) wire enrichment into internal tools on a fixed per-record shape. Journalists and academics (2) get dated, attributable facts for the startup-and-industry story - see journalists academics use cases.

Within industrial conglomerates data, Crunchbase owns private-market events and concedes nearly everything else. SEC EDGAR Company Facts & Submissions API scores 10/10 in our catalog and carries standardized XBRL facts for every US-listed filer - GE alone exposes about 945 us-gaap concepts, with submissions history back to 1994 - but holds nothing on private rounds, because private companies do not file. Companies House Register Search (UK) traces officers and ownership across roughly 5 million live UK companies including private ones, but stays register-shaped and UK-bound. Fortune Global 500 ranks the world's largest companies by revenue once a year - a scale benchmark, not an event stream.

What none of them replicates: funding rounds, investor networks, acquisition events and predictive signals for companies that exist outside the filings-and-registers world. The usual pattern is to stack them - EDGAR and Companies House for structure and financials, Crunchbase for the private-market narrative between filings. The full slate sits on our best industrial-conglomerates datasets ranking.

Three things worth settling upfront. First, schema provenance: the operator publishes no field-level dictionary, so the ten-field structure above traces to observable surfaces and gets pinned to exact column names and types against your delivered extract before anything downstream depends on it. Second, scope discipline: millions of profiles reward a named universe - bring your watchlist, your sector tags or your competitor set and the sample arrives cut to it. Third, treat the predictive scores as what they are - modeled probabilities riding beside the records, useful as triage, not as verdicts. Get a sample of this dataset scoped to the companies you actually track.