Dark Mode
Foundation Intelligence
Foundation Model Datasets
Tags and Keywords
"No reviews yet"
£15,000
About
Vivameda - Longitudinal Company Evolution Panel
A 70-year longitudinal panel of company workforce evolution. Built as training substrate for AI models that reason about how companies scale, fail, and recover.
Overview
- 4.2M companies across 100+ countries
- 48M observed company-year records, 1950–2020
- 1.88B skill rows, 46.5M capability buckets
- One row per company per year. Year-end resolution.
- Real historical observation, not synthetic, not scraped from the web.
Why It Exists
Modern AI models reason about companies almost entirely from post-2020 web data. They never saw the dot-com bust, the 2008 recovery, or the full seventy-year arc of how organizational structure predicts outcomes.
That's not a knowledge gap. It's a substrate problem.
Vivameda is the substrate.
Era Coverage
1950–1969: Post-war industrial expansion
1970–1979: Stagflation, oil shocks
1980–1989: Deregulation, conglomerate era
1990–1999: Globalization, early internet
2000–2009: Dot-com cycle, 2008 crisis
2010–2017: Platform economy, mobile wave
2018–2020: Pre-COVID dense window (highest observation density)
Product Tiers
Workforce Intelligence (Entry) — Base panel, 19 columns
Landscape Intelligence (Mid) — Workforce + market structure, 24 columns
Workforce and Capability Intelligence (Premium) — Full schema, 30 columns at company-year-bucket grain
Use Cases
- Foundation model pre-training on company evolution patterns
- Fine-tuning corpora for company-reasoning systems
- Evaluation benchmarks for organizational reasoning
- Ground-truth anchoring for agent systems
- Alternative data signal for systematic strategies
Delivery
- CSV / Parquet flat files (one-time license)
- Snowflake Data Sharing (recurring access)
- Custom delivery formats on request
Data Provenance
Aggregated at the company-year level from commercial partnerships and publicly available workforce signals collected over the past decade. No PII. No individual employee records. All observations aggregated to the company-year grain.
Note on Recency
The dataset terminates at 2020 by design. This is a feature: it provides a stable, reproducible historical training corpus that does not drift between training runs.
Listing Stats
VIEWS
7
DELIVERY
INSTANT DOWNLOAD
LISTED
21/04/2026
UPDATED
27/04/2026
REGION
GLOBAL
QUALITY
5 / 5
Loading...
£15,000
Download Dataset in Parquet Format
Recommended Datasets
Loading recommendations...