The prompt
Score 10 of 10, Excellent
Use it when A table, export or shared spreadsheet is used by several people and you need every column documented so everyone reads it the same way.
- {{dataset_name}}
- {{readers}}
- {{sample}}
- {{known_facts}}
Full prompt and why it works
Prompt
Write a data dictionary for the {{dataset_name}} data below, for {{readers}}, so anyone using it knows what each column means, what values to expect and where the traps are, without asking the person who built it. Column names and a sample of rows (20 to 50 rows is enough): """ {{sample}} """ What I know about how the data is created (source system, who enters it, when it refreshes, known quirks): """ {{known_facts}} """ Produce: 1. Overview: what one row represents, the source, the refresh schedule and the date range, in up to four sentences. 2. A table with one row per column and the columns Column, Meaning, Data type, Example value (from the sample), Allowed values or range, Can be blank (yes or no), Notes. 3. Keys and joins: which column or columns identify a row uniquely, and which link to other tables, where my facts say so. 4. Known issues: blanks, mixed formats, codes without labels or values that look wrong in the sample, each with the column and an example. 5. Questions for the data owner: columns whose meaning needs confirming. Write meanings a new colleague could act on, for example "Date the order left our warehouse, in local time; blank until it ships" rather than "Ship date". This example shows the form only. Base data types, examples and ranges on the sample, and say "in the sample" where the full data could hold a wider range. Take meanings from my facts. Where a column's meaning is not stated and its name and values do not make it clear, write "Unclear:" followed by your best reading and add the column to the questions, instead of presenting a guess as a definition. Do not copy personal data from the sample into the dictionary; mask names, emails and phone numbers in example values. Keep each Notes cell under 25 words.
Variables
- {{dataset_name}}
- The name people use for the data, for example "orders export" or "the customer master sheet".
- {{readers}}
- Who will use the dictionary, for example "analysts and the finance team".
- {{sample}}
- The header row and 20 to 50 rows of data, including some messy ones.
- {{known_facts}}
- Where the data comes from, who enters it, when it updates, and any quirks you know.
Why it works
- What good looks like. Meanings must be something a new colleague could act on, so "Ship date" becomes when, where and when it is blank.
- If unsure. Columns the facts do not explain are marked "Unclear:" and become questions for the owner, never confident definitions.
- What to avoid. Personal data in the sample is masked, so the dictionary can be shared more widely than the data itself.