Preparing source files for conversion.

Get the shape right

A source file converts cleanly when it is a plain rectangle of data. That means one table per file, a single header row, one fact per cell, and no decoration.

• No merged cells. A merged cell has one value and several apparent positions, so the reading is ambiguous by construction.
• No blank spacer rows or columns inside the data region. They terminate the table for most readers.
• No subtotals mixed into the detail. A subtotal row that looks like a data row will be read as one, and the totals then double count.
• No notes in data cells. Commentary belongs in its own column or its own file.
• One report per file where the framework treats them as separate tables, even if your system exports them together.

Headers and identifiers

Headers are how a mapping stays attached to your export from one period to the next. Three rules keep that link intact:

1. Stable header text. A renamed column is an unmapped column. If a system rename is unavoidable, the mapping is updated deliberately rather than discovered at validation.
2. Unique headers. Two columns with the same name cannot be distinguished, and the second is usually the one silently dropped.
3. Stable row identifiers for tables that take an open list of rows — counterparties, instruments, properties. Regenerating identifiers each period makes cross-period comparison impossible even when the data is correct.

Values, scale and signs

Encoding and characters

Export as UTF-8. Legacy encodings corrupt accented characters in counterparty and entity names, and the corruption often survives validation because the value is still a valid string. Characters that are reserved in structured formats need encoding rather than removal — stripping them silently changes a legal name.

What not to fix in a spreadsheet

The instinct to correct the file rather than the source is what makes each period as expensive as the last.

• A wrong value belongs in the system of record, or the same finding returns next period.
• A vocabulary mismatch — your value against the framework’s coded list — belongs in the mapping, where it clears every occurrence at once.
• A structural problem belongs in the export definition, not in a hand-reshaped copy.
• Nothing belongs in the output package. Edits made after generation are lost on the next run and break the link between the filing and its validation record.

Pre-upload checklist

1. One table, one header row, no merged cells, no spacer rows, no subtotals in the detail.
2. Headers unchanged from the previous period, or the mapping updated deliberately.
3. Row identifiers stable for any open table.
4. Scale and currency stated and consistent across the file.
5. Numeric columns stored as numbers; dates in one format.
6. Blanks left blank.
7. UTF-8 encoding.
8. Reporting entity, level and reference date confirmed before upload.

REGREP is an independent software provider. This record explains a reporting framework in plain language and is not legal, tax or regulatory advice. Confirm scope, thresholds and submission dates with your competent authority before you file.