Field note
How can I review data-dictionary clarity?
Ask someone who did not write the dictionary to explain selected fields, coded values, missing states, and timing in their own words. Record confusion as revision evidence rather than blaming the reader.
Choose representative fields
Pick high-impact fields, easily confused codes, measures with units, and fields with missing states. A random sample alone may miss the riskiest interpretation problems.
Ask open questions
Ask the reader to explain meaning, timing, allowed codes, and missing states without quoting the definition. Open questions reveal ambiguity better than yes-or-no approval.
Treat one result as bounded evidence
Record who read it, the question you asked, the outcome, and a note on what you changed as a result. The record is evidence that one conversation happened with one reader, which is exactly as much as it claims. A clear result on one field, with one reader, does not certify the dictionary for every audience, and a second reader from a different team often finds a different gap.
Worked example from the sample project
The sample ships with two fields already marked needs-review, condition and access_note, and they are the right two to take to a reader. Condition is a bounded category whose thresholds live in an observer's head. Access_note is free text, which means its content is whatever a field worker decided to write on the day.
Ask a reader who did not write the dictionary to explain, in their own words, what watch means and when they would expect to see it. If they answer with the word watch, or with a paraphrase of the label, the definition has not done its job. Record what they actually said, and record it as evidence for a revision rather than as a complaint about the reader.
What this does not do
- One clear review result does not certify a field for every audience.
- The studio does not score comprehension or grade a reader.
- A recorded review is evidence of a conversation, not of organisational approval.
Do this in the studio
Choose the fields most likely to be misread, not a random sample.
- Open the Reader review station and pick coded fields, measures with units and fields with missing states.
- Ask open questions and write down the reader's own words.
- Mark the result and revise the folio before re-running the deterministic review.
Open the studio at the Reader review station