Practical digital skills

Assess a dataset for your question

Check definitions, coverage, missing values and provenance before drawing conclusions from a table.

Step-by-step guideUpdated
Follow the method ↓
A person checks a dataset's definitions and missing values

The answer in 30 seconds

A clean-looking table can contain duplicates or missing fields. Quality depends on the question: the same file may support a list of places but fail to support a fair comparison between regions.

Keep the original and documentation, define what one row means, inspect essential missing values and units, then state which uses the data does and does not support.

Examples to adapt

A blank capacity field

Missing does not mean zero; avoid a calculation that silently makes that assumption.

Two matching names

They may be distinct locations. Test duplicates using the fields that define an observation.

An old data release

Check whether its time period fits the decision and whether a newer release exists.

Follow the method

  1. 1
    Preserve provenance

    Save the original file, version, publisher and documentation. Work on a copy.

  2. 2
    Define one row

    Write down whether it represents a person, place, event or period.

  3. 3
    Check completeness

    Count missing values in fields required for your question; investigate special codes.

  4. 4
    Check consistency

    Review units, dates, categories and possible duplicates using the correct key.

  5. 5
    State a limited conclusion

    Say what the data can answer, what remains uncertain and which transformations you made.

A checklist to keep

Use these checks to record what you found. The grid supports a decision; it does not make one for you.

Assess a dataset for your question: checklist
CheckWhat to examineAction
UnitWhat does one row represent?Define
CoverageWhich places and dates are included?Check
MissingWhat do blank or special values mean?Count
ConsistencyAre units and identifiers coherent?Inspect
ProvenanceCan you reproduce each change?Record

Download the CSV checklist

Free, no sign-up. UTF-8 text with semicolon-separated columns.

What to check

An unusual value needs investigation, not automatic deletion.

A fully populated column can still contain the wrong unit.

Corrections should be traceable and reversible.

Common questions

Is checking a sample enough?

It helps you understand the file, but full-column checks are needed for missing values, types and duplicates.

Can I publish a corrected copy?

Check reuse terms and document changes clearly without implying the producer approved them.

Sources and documentation

Documentation consulted on . Examples are illustrative; interfaces and results may change.

Continue exploring

All English guides →