What does a missing or blank code mean?

A missing or blank code means that no code was published in the place where the code was expected. It does not prove that the unit has no activity, that it is not registered, or that no code is held for it elsewhere. A blank can mean that the code was suppressed, that the record was incomplete, that the field does not apply to the unit, that the dataset does not publish codes at that level, or that the copy from which the blank was taken omitted the field.

Each of those causes has a different implication, and the cause must be found before the blank is interpreted. A blank quoted as a finding, without its cause, is an interpretation without evidence, and it should be withdrawn.

Why are codes suppressed or omitted in datasets?

Codes are suppressed or omitted in datasets for reasons that belong to the dataset, not to the unit. A cell in a published table can be suppressed where publishing it would disclose information about a single unit, or where the cell is based on too few units to be reliable. A dataset can omit codes at class level and publish only at section or division level, where the detailed breakdown is not required or is not reliable. A dataset can omit a code for a unit where the unit’s activity description was incomplete at the time of compilation, and the code was not assigned. Eurostat describes, on its transition page, the production of parallel datasets and backcast series, and a dataset in transition can omit a code where the double coding for a unit is not yet complete, or where the correspondence for a split class has not yet been applied. None of those causes is a statement about the unit. All of them are statements about the dataset, and they should be recorded as such, with the dataset metadata cited.

Why are codes blank in listings and registers?

Codes are blank in listings and registers where the record is incomplete, where the field does not apply, or where the copy is out of date. A listing that copies a register at intervals can show a blank where the register record was incomplete at the copy date, even though the register has since been completed. A register record can show a blank where the unit type does not carry an activity code in that register, or where the code is held in a different register, such as a statistical business register rather than a business register. The Central Statistics Office states that its Business Register completed updating the coding of enterprises to NACE Rev. 2.1 during 2025 (FIXED, as stated by the Central Statistics Office), which implies that, before that completion, some enterprise codings in that register were still on the earlier basis or incomplete. A blank in a record made before a recoding was completed should therefore be read as a record in progress, not as a permanent absence, and it should be rechecked after the recoding, with the new check date recorded.

How is a blank investigated?

A blank is investigated by checking the source, the field, and the holder, in that order. First, check the source: is the blank in a dataset, a listing, or a register, and what does the source state about suppression, omission, or incomplete records? Second, check the field: is the field for a principal code, a secondary code, or a code at a level the source does not publish? Third, check the holder: does the register that holds the record show a code for the unit, using the unit identifier? The table below shows the investigation as a record, with the cause row left for the finding, so that a blank is never recorded without its cause being sought.

Step	Question	Record
1	What is the source?	Dataset, listing, or register, named
2	What does the source state about blanks?	Suppression, omission, or incomplete record, as stated
3	Which field is blank?	Principal code, secondary code, or level not published
4	What does the holder show?	Register entry, with identifier, code, version and date
5	What is the cause?	Cause found, or cause not found as checked on 28 October 2026

The last row records the outcome honestly. Where the cause is not found, the record states that the cause was not found as checked on 28 October 2026, and the blank remains unexplained, rather than being given a cause by assumption.

What does a blank not prove?

A blank does not prove that a unit has no activity, that it is dormant, that it is not registered, or that it has no code in any register. It proves only that the place consulted shows no code. A unit can be active with a code held in a different register, a unit can be registered with a code that a listing omitted, and a unit can have a code in a dataset at a higher level while the class level cell is blank. The EUR-Lex summary states that the NACE legislation only applies to the use of the classification for statistical purposes, and does not oblige member states to collect, publish or supply data, which means that a blank in a published statistic can reflect a decision not to publish, not a fact about the unit. A blank should never be quoted as evidence about a unit, and a blank in a listing should never be quoted as evidence that the register holds no code.

Where is a blank resolved?

A blank is resolved in the register that holds the record, using the unit identifier, and in the dataset metadata, for the suppression or omission rule that applies. The meaning of a category, where a code is found, is confirmed in the Eurostat explanatory notes. A blank that cannot be resolved, because the holder cannot be opened or the metadata does not state a rule, should be recorded as unresolved, with the date of the check, and the record should be rechecked later, rather than being filled with an assumed code or an assumed cause.

Investigate the source, the field and the holder, record the cause or record that it was not found, and confirm the entry in the register that holds it.