Data sources and registers
Annual accounts as data: making registers machine-readable
Starting questionHow do published annual accounts of German companies become comparable figures that can be analyzed reliably?
Result
Works
The local mirror delivers comparable figures in the HGB structure, with a plausibility check and provenance for every document.
Through a local mirror: a company’s publications are copied once in full; after that, only new ones are synchronized. Every document exists twice, true to the original and normalized to the HGB structure, with a plausibility check and provenance. Accounts from 2022 on are in the Unternehmensregister.
What we built
We built a server that mirrors all publications of a company into a local database once and from then on only reconciles what is new.
Every document exists twice: true to the original and normalized to the German commercial code (HGB) structure. On top come a plausibility check (total assets equal total equity and liabilities) and a provenance record with a hash.
What it means for business architecture
External data only becomes an asset once its provenance, completeness, and coverage are proven. An answer must also say which data is missing.
Lessons learned
- Accounts from 2022 on sit in the company register (Unternehmensregister), no longer in the Federal Gazette. Miss that boundary and you wrongly report “no data”.
- When a portal offers no stable addresses, a mirror is the only stable foundation.
- Keeping raw and normalized data apart makes the analysis auditable.
- annual accounts
- company register
- data provenance
- HGB