Work demonstrated

DORA register lab: two corrections, one unresolved finding

We ran four local checks on deliberately flawed synthetic data, corrected two confirmed defects and repeated the checks. One missing identifier remained unresolved. You can download the original files, program and actual results and reproduce the exercise.

Executed 14 September 2026 · laboratory data · tool 1.0.0

What we actually tested

Two small UTF-8 CSV files using the educational format lab-csv-projection-v1. Three selected fields correspond to contract references and a using entity’s identifier in ITS 2024/2956. This is not an official xBRL-CSV filing package or a complete regulatory template.

  1. A master contract reference is not empty.
  2. A master contract reference is not duplicated.
  3. Each usage reference matches a contract in the master file.
  4. The using entity’s identifier is not empty.

Observed before and after

FindingAction supported by the synthetic source recordRetest
Duplicate LAB-001 master rowRemove the confirmed exact duplicateResolved
Usage reference LAB-01 has no matching contractCorrect to LAB-001 after explicit confirmationResolved
LAB-002 has no using-entity identifierRequest the missing fact; leave it unresolvedStill missing

The result is three findings before correction and one afterwards. “Corrected” does not mean “complete”: copying another entity’s identifier would conceal the problem. The change record states what was changed, why and what still needs a data owner’s answer.

Inspect and reproduce the work

The package contains original and corrected CSV files, the synthetic source record, field-level findings, ten executed counterexample tests, runtime details and SHA-256 hashes. Two local runs produced byte-identical reports. Python 3 and its standard library are sufficient; the program makes no network requests.

Download files, program and evidence (ZIP)

Read the actual retest (JSON)

What this does not establish

The four checks are a deliberately narrow subset. A nonempty invalid LEI and duplicate usage rows pass this program; the tests expose those limits. It does not validate identifier format or status, all table keys, enumerations, dates, the full taxonomy, business facts or the reporting package. Nothing was submitted to an authority, and there is no supervisory acceptance or customer result.

Basis and version

The selected relationships refer to Annex I/II of Commission Implementing Regulation (EU) 2024/2956 as published on 2 December 2024. LAB-R rule labels are our local labels, not EBA validation codes. A real engagement needs its own supported reporting format, version and complete check scope.

Have a returned register?

We first review the actual package and error messages, then confirm whether the format and required checks can be supported. This laboratory demonstrates traceable correction work; it does not promise support for every reporting package.

Discuss a register-only review