A polished report can contain weak data. The useful question is whether a reader can reconstruct the information, not how many pages were produced.
What the study does and does not establish
CNDCEC IRS 19, November 2025, provides a descriptive review of early VSME reporting. Its introduction refers to forty reports, while the body also says more than forty; precise percentages should not be inferred.
It identifies progress and limitations, including generic descriptions and methodological weaknesses. It neither certifies the reports nor ranks every platform in the market. Sample observations do not establish that every automatically generated report is unreliable.
Our pre-submission checklist
These are VSME-PRO editorial suggestions, not a reproduction of CNDCEC's methodology.
Reconstruction test. Ask a colleague to recover a value's source, period, unit and calculation without an oral explanation.
Specificity test. Read a paragraph without the company name. If it could describe anyone, add relevant facts or remove it.
Consistency test. Compare repeated totals and check that the same terms use the same definitions.
Gap test. Examine zeros, empty cells and “not applicable”. Verify that each really represents that status.
Recipient test. Check that the report answers the request without exposing unnecessary attachments.
A synthetic example
“The company continually invests in people” does not explain an activity. Aggregated training hours with period, coverage and source can be checked. If unavailable, identify the gap and action needed; do not invent a result.
Where software helps
Software can organise work and support checks, but cannot by itself establish that a source is correct. Review remains necessary even when a file passes technical validation.
Start with the data inventory. To distinguish digital structure from substantive quality, see the XBRL guide.