Use cases & endpoints
CVM CDA fund portfolio data (Brazilian holdings)
CDA is the Brazilian equivalent of a holdings disclosure, and it is unusually detailed: fund portfolios reported to CVM with positions, weights and asset classes, published as monthly archives keyed by competência. Compared to a quarterly 13F it is a generous disclosure regime.
The friction is entirely mechanical. Archives are zipped, encodings are inconsistent across older years, asset-class labels are Portuguese, schemas have shifted over time, and the file describes a month that closed before it was published.
Hindcast normalizes CDA into a positions model that matches the US holdings schema, resolves instruments and issuers into the shared entity graph, and carries the disclosure date so point-in-time queries behave correctly. That turns a familiar exercise — download, unzip, transcode, reconcile — into a query, and makes Brazilian holdings comparable to US ones without a manual mapping layer.
hindcast — session
previewhindcast> positions and weights for a fund in a given month
→ tool: cvm_fund_portfolio { cnpj:"...", competencia:"..." }
✓ positions with weights, asset classes and disclosure date
source: CVM CDA archive for that competência
What's covered
- Positions with weights and asset classes by CNPJ and competência
- Portuguese asset-class labels mapped with originals retained
- Disclosure date stored separately from the reporting period
- Instruments and issuers resolved through the entity graph
- Historical schema variations normalized rather than dropped
- Same positions model as US 13F holdings for direct comparison
Join the waitlist — this page's dataset is already selected.
Questions
How does CDA compare to a 13F?
It is more frequent and more detailed on asset classes, but it covers a different universe under different rules. The schemas align on positions; the regime differences stay visible as fields rather than being smoothed away.
Why is the disclosure date emphasized?
Because the portfolio describes a month that had already ended when the file was published. Treating the period end as the knowable date is the standard way to accidentally give a backtest several weeks of foresight.
Related
Join the waitlist. Vote the roadmap. First in, first served.