Import formats and canonical headers

CSV, Excel, and ODS

Use UTF-8 CSV, .xlsx, or value-only .ods. Legacy binary .xls is not accepted; convert it in a trusted spreadsheet application, inspect the converted workbook, and save it as .xlsx. A canonical closed-table row contains table_code, row_code, column_code, and value. A canonical cell identity can instead use cell_id and value. Open tables and tables with additional axes require the dimensions needed to identify a unique cell:v2 context.

Example synthetic CSV:

table_code,row_code,column_code,value

S.01.02.01.01,R0010,C0010,Synthetic value

Column names must be unique. Duplicate logical cells, unknown templates, invalid values, and noncanonical aliases are rejected. CSV follows quoted-field rules, including escaped double quotes.

Decimal text from CSV and ODS remains decimal text through mapping, preview, validation, and apply. Scaling, offsets, sums, tolerance checks, and reconciliation use decimal arithmetic and do not route reported values through JavaScript binary floating point.

Generated filing template

Select Download Excel template in a filing's import panel to create the recommended .xlsx starting point. The workbook is specific to that filing and contains an instruction sheet plus one worksheet for each applicable EIOPA form. If the regulatory S.01.01 contents template explicitly marks another form as reported, that form is included too, so the workbook never declares a reported form while omitting its worksheet. Any disagreement with the filing's confirmed applicability decisions remains visible to validation and must be resolved before review or export.

The Instructions sheet includes Workbook contents, with the exact regulatory code, the full localized form name, and an Open sheet link for every included form. Visible worksheet tabs use the exact regulatory code only, so the full name is not cut off by Excel's worksheet-name limit. Use the links to navigate; do not rename the worksheets.

Fixed forms use a reporting matrix: reporting items run down the rows and reporting columns run across the worksheet. Repeating forms use an Excel table with one asset, counterparty, policy, or other record per row and one reporting field per column. They include 25 ready-to-use rows, and additional records can be added as new Excel table rows.

Enter or paste data only in highlighted cells. Input-cell notes describe the expected format and requirement status. Every bounded value uses a dropdown backed by a hidden lookup range, including long lists and values containing punctuation.

The Required column preserves three states:

• Yes means the cell is explicitly mandatory.
• No means the cell is explicitly optional.
• Depends on filing means the taxonomy does not define a universal cell-level requirement. Use the filing's applicability decisions, validation results, and reporting guidance. Do not interpret this state as optional.

Canonical cell, table, row, column, and repeating-record identifiers are kept in hidden technical content so the workbook can be imported without a custom mapping. Forms with additional contextual axes use a normalized fallback to preserve exact contextual cell identities. Do not rename worksheets, remove hidden content, or move worksheets between generated workbooks. Metadata also binds the workbook to the filing identity, taxonomy checksum, reporting period, and template format version; a template generated for another filing is rejected.

Blank value cells are ignored and do not clear existing filing values. To replace or clear existing data, use the filing controls and verify the complete preview rather than relying on blank spreadsheet cells. Formula cells are not calculated during import and must be replaced with reviewed values.

The complete taxonomy workbook set is a reference and testing library. Normal preparation should use the smaller filing-specific download so the workbook contains only applicable forms.

Mapped tabular sources

A mapping JSON may translate customer headers and supply supported defaults for CSV, Excel, or ODS. One mapping can include several sheets. In addition to normalized sources with one reporting value per row, mappings can reshape:

• Rows as records: one asset, counterparty, policy, or other entity per source row, with reporting fields across columns.
• Columns as records: one entity per source column, with reporting fields down rows.

The first preview detects these layouts from exact taxonomy codes and labels. Review each proposed sheet, correct its key, fields, destination codes, and context dimensions, then inspect the transformed sample. Save the combined mapping, select it, and preview again. The saved mapping performs a deterministic reshape into canonical reporting facts and preserves the original workbook cell address in lineage. Contextual tables require every dimension exactly once, using a source field, the record key, or a reviewed fixed member.

Suggest mapping with AI is an optional second pass. The organization must be allowed to use it. The operator chooses one of three data-sharing modes:

• sheet names and headers only;
• headers plus structural value types, without raw sample values; or
• headers plus at most five raw sample rows, when organization policy also allows value sharing.

Its structured response is constrained to the previewed source fields and the filing's pinned taxonomy codes. Suggestions below the configured confidence floor and invalid references are discarded. A retained proposal records the provider, model, prompt-contract version, privacy mode, and suggestion time. It cannot change filing data: a user must save it, select the saved revision, run a fresh preview, and explicitly apply the validated changes.

Preview the complete change set. Saved mappings are organization-scoped and reloaded when apply runs. Administrators can compare a JSON draft to the selected revision, test it against the current source, save it as the next revision with stale-write protection, or clone it under a new identity.

ODS safety and limitations

ODS imports read worksheet values and supported scalar types from content.xml. Lexical decimal values are preserved rather than converted to JavaScript numbers. Excel numeric cells may already contain binary floating-point values; when every source digit must survive, store the value as text in Excel or use CSV/ODS decimal text and confirm the preview. Formulas are reported and refused rather than executed or replaced by cached results. XML entities and document-type declarations are refused. Compressed content, repeated rows and columns, expanded row and column counts, and materialized sheets are bounded. Macros, external calculations, styling, charts, and other document features are not import inputs.

EIOPA XBRL

The instance must match the filing’s pinned supported taxonomy, entity, period, filing indicators, contexts, units, and dimensions. Parser and size limits fail closed.

National XBRL

Only profiles explicitly described as supporting national XBRL import are accepted. Profile identity, schema references, contexts, facts, and units are validated against implemented rules.

Safety rule

Always preview before apply and reconcile after apply. File-format acceptance does not establish regulatory or business correctness.