Report or Dataset First? How SMEs Should Implement Voluntary Sustainability Reporting

A practical implementation guide covering users, boundaries, evidence, version control, output formats and the role of the 2026 EU Voluntary Standard.

A voluntary sustainability project often starts with a request to “prepare the report”. That language encourages the team to open a document, choose a design and begin writing. The problem is that the same SME may subsequently receive a bank questionnaire, a customer portal request, a tender schedule and an internal management request for the same information in different formats. If the only controlled object is the finished PDF, every new request becomes a manual extraction exercise and every change risks creating inconsistent answers.

The 2026 Voluntary Standard was designed for more than one audience. Its objectives include meeting the information needs of value-chain partners, banks and investors, improving the undertaking’s own management of sustainability issues and supporting resilience and access to finance. The Standard also states that the report’s primary function is to inform actual or potential business counterparties; public publication is an option rather than the sole purpose. This multi-user design makes a reusable dataset particularly valuable.

Figure 1. One controlled dataset can produce several outputs without rebuilding the underlying evidence.

Dataset-first does not mean collecting every possible ESG datapoint. It means identifying the legitimate users and then creating one governed source for the information that the organisation decides to report. A user map should distinguish the information need, the permitted use, the timing, the level of detail and whether the request is covered by the value chain cap or arises from another legal or contractual basis.

A datapoint is not controlled unless the team knows what it covers. B1 requires the undertaking to state whether the report is prepared on an individual or consolidated basis and, for a consolidated report, to list the subsidiaries covered. The reporting period should be consistent with the financial statements where those are prepared. Beyond that overall basis, individual metrics may require additional operational boundaries, such as the sites included in energy data, the employee population used for workforce metrics or the facilities assessed for biodiversity sensitivity.

The unit of control should normally be the paragraph or subpoint, not the page on which it will eventually appear. The Standard contains essential datapoints, items reported only in specified circumstances, explicitly voluntary datapoints and items that are considered only for sector information. Some environmental datapoints are also voluntary for undertakings with 10 employees or fewer. A data dictionary should therefore record the regulatory category and the organisation’s selection before data collection begins.

Paragraph 11 requires information to be relevant, faithful, comparable, understandable and verifiable. Verifiability is difficult to achieve when evidence is assembled only after the narrative has been approved. Every quantitative value and material narrative claim should therefore carry a source reference, owner, methodology, review status and retention location. The evidence itself may be restricted; the register can still record its existence and access level.

Quantitative metrics: source-system extract, invoice or register; calculation workbook; conversion factors; boundary reconciliation; reviewer evidence.

Policy disclosures: approved policy, version, effective date, scope, responsible owner, evidence of approval and public-availability status.

Initiatives and targets: authorised plan, baseline, boundary, target date, progress evidence and governance approval.

Conditions and non-applicability: documented facts showing why a paragraph-level trigger was or was not met.

Output claims: evidence that the published wording is consistent with the approved dataset and does not imply a wider scope.

The reporting dataset should have a period version and a release status. Changes after approval should be recorded with the reason, preparer, reviewer, date and affected outputs. From the second reporting year, the Standard requires comparative information for the previous year except for metrics disclosed for the first time. A roll-forward process should therefore preserve the prior-year approved value, distinguish restatements from current-year changes and explain methodology changes.

Figure 2. Dataset-first implementation lifecycle and annual control loop.

The Standard does not force every undertaking to publish a glossy standalone report. The primary function is to inform counterparties; an undertaking may make the report public, place it in a management-report section or issue a separate document. It may also incorporate information by reference where the referenced documents are accessible at the same time and from the same document set. The correct output is therefore a design decision based on users, not a substitute for the selected reporting option and its datapoints.

EFRAG’s June 2026 VSME Digital Template version 1.3.0 supports data entry, checks, paragraph links and conversion to Inline XBRL. Its XBRL taxonomy provides a vendor-independent data model and supports structured exchange. Those features reinforce the value of data-first reporting. However, the June 2026 materials expressly reflect the 2025 VSME Recommendation for SMEs with fewer than 250 employees, and EFRAG indicated that an update would be needed when the future delegated act was released.

The explanatory note also identifies practical limitations: the template does not support comparative roll-forward, does not collect or calculate all underlying data, has no GHG calculator, can be corrupted through spreadsheet modification and may be suitable mainly for first-time reporting. A sensible approach is to maintain the authoritative dataset and evidence register independently, then populate the current template or digital interface as an output once its mapping to the applicable 2026 Standard has been checked.

Map users and purposes. Record who asks for information, why, when and in what format. Separate CSRD value-chain requests from other legal or contractual requests.

Select the reporting option. Decide whether the organisation will prepare the Basic module only, pursue the Basic and full Comprehensive compliance option, or provide selected additional Comprehensive disclosures without representing a full Option B report.

Lock the reporting basis and period. Record individual/consolidated basis, subsidiaries, period and metric-specific boundaries.

Create the datapoint register. Map each paragraph/subpoint, category, unit, condition, owner and output requirement.

Build the evidence register. Link each datapoint to source, methodology, assumptions, reviewer and confidentiality status.

Collect, calculate and classify. Distinguish reported values, estimates, not-applicable conditions, voluntary non-selections, permitted omissions and unresolved gaps.

Review and freeze the annual dataset. Perform reconciliations, consistency checks, management review and approval before design.

Render each output. Use controlled mappings for the report, customer pack, lender response and digital file; label scope and status.

Roll forward. Preserve comparatives, record restatements and update only changed facts, methods, evidence and conditions.

For an SME, the most valuable first-year asset is not necessarily a polished report. It is a controlled, reusable and reviewable sustainability dataset that can support the selected Voluntary Standard option and be rendered for different users. The public report still matters: it explains context and makes the information understandable. But when the report is generated from a governed dataset rather than assembled as a one-off publication, the organisation gains consistency, faster responses, clearer evidence and a much stronger starting point for the second reporting year.