Data and spreadsheets
Make a data dictionary for a small project
A data dictionary lists every field with one agreed meaning, a data type, allowed values and what a blank means. Build it from the real column headers and a few sample rows, then settle any field where two people would describe it differently.
A text to start with
My project tracks [what the data records]. The export has these columns: [column names]. Here are a few anonymous sample rows: [rows]. Draft a data dictionary with a definition, type, allowed values, meaning of a blank and source for each field. Mark any definition you had to guess. Ask me about fields whose meaning is unclear instead of inventing one.
Example request. Change the details to fit your situation.Data dictionary sheet
Project and data file: [ ] One row represents: [ ] For each field, copy the block below: Field name: [ ] Plain definition: [ ] Type (text, number, date, yes/no): [ ] Allowed values or format: [ ] What a blank means: [ ] Where the value comes from: [ ] Who can change it: [ ] Similar field it could be confused with: [ ] How the two differ: [ ] Last reviewed on: [ ] Open questions: [ ]
Fields that blur together
Similar column names often hide different concepts, such as the date an item was created and the date it was last changed, or the person who requested something and the person who completed it. Check that each definition names the event or person it refers to. A blank can mean unknown, not applicable or not yet entered, and those should not share one meaning. Look at allowed values for spelling variants and for codes that mean different things in different exports. If a definition could describe two columns equally well, it needs sharpening.
Ask another agent to check the result
Review this data dictionary and flag at least two fields that look similar but represent different concepts, such as created versus updated dates or requester versus completer. For each pair, quote both definitions, explain how a reader could mix them up, and suggest clearer wording based only on the sample rows.
Settle the disputed fields
Mark every field where the draft guessed, and ask whoever enters or exports the data to confirm each one. Decide what to do with values outside the allowed list, and who may add a new field. Save the dictionary beside the data file and record the date of the last review. Update it whenever an export changes.