Counts show all available guides. Filters combine to narrow your results.
Build a workflowGuide
Explain what every column in a small dataset means
A column-by-column meaning sheet with types, units, blank rules and the rule that identifies each row.
Illustrative resourceNot tool-testedAdults working with original, synthetic or authorized non-sensitive material
What you’re working towards
A data dictionary with row grain, units, allowed values, missing markers and questions that stop unsafe interpretation. This is an authored practice example; no consumer AI run or real participant trial has been recorded.
Your choice also saves this guide’s current worked-example checkmarks. Only this browser; no account or sync. View saved guides
Saving needs JavaScript. Reading and downloading this guide does not save it.
Make a start
One step at a time.
2 steps
Start with the synthetic worked example or other non-sensitive inputs. Run one prompt at a time in your chosen AI tool. Replace any bracketed placeholders with the requested material, check each reply, and supply the source or earlier outputs that step needs. Paste them again if the tool lacks that context.
Start with the synthetic example, then adapt each prompt to your task.
01
Try the complete example
Open a new chat in a text AI assistant you already have permission to use. Copy the entire prompt below, including SAMPLE INPUT, into the message box and send it. No upload or special tool is needed. Expect a column-by-column meaning sheet with types, units, blank rules and the rule that identifies each row. Compare the result with Input; Output is a worked illustration, not text you need to paste. There are definitions for all four exact column names and one two-column uniqueness rule. P01’s zero and P02’s blank remain distinct; the negative test row is rejected. A repeated item on a different date is allowed; a repeated item/date pair is not. If a detail differs, paste: “This part is wrong: [quote it]. My input says: [quote the exact source line or rule]. Return a complete corrected replacement and explain the change.” Recheck it yourself. If the source is missing or contradictory, leave that part unresolved and obtain the needed fact before using it.
Your starting prompt
Write a complete data dictionary from the owner definitions. For every column give meaning, type, unit, valid values, missing rule and example. Explain the row meaning and composite key in plain words. Apply the rules to both source rows and the separate negative test row, without changing or silently repairing any data.
SAMPLE INPUT
Original synthetic practice input. The people, organizations, records and events below are fictional.
Complete fictional paper-count table. Vertical bars separate fields; the empty qty field below is intentionally blank.
item_id | counted_on | qty | unit
P01 | 2026-10-01 | 0 | sheets
P02 | 2026-10-01 | | sheets
Owner definitions:
- One row is one item’s count record for one calendar date; the record may say it was not counted.
- item_id is a nonempty text identifier. Preserve its characters and leading zeros. No further identifier pattern is supplied.
- counted_on is a required valid calendar date written YYYY-MM-DD; no clock time or time zone is stored.
- qty is either a nonnegative whole number or empty. Zero means counted and none observed. Empty means not counted. Negative numbers, fractions and text are invalid. No maximum is supplied.
- unit is required and must be the exact text sheets in this table.
- item_id plus counted_on together must be unique; the same item may occur on a different date. This two-field combination is called a composite key.
Also explain this separate deliberately invalid test row, which is not part of the two-row source: P03 | 2026-10-01 | -1 | sheets. Do not clean or change the source.
Return written text only. Keep this fictional and label illustrative dialogue as invented. Do not claim an actual interview, survey, learner trial or tool execution. Do not send, publish, collect responses or change another service.
02
Use your own input
Open a new chat. Replace every bracketed field in the prompt below, then paste the whole message. Change related names, source IDs, counts, dates, labels, limits and checks together; none of the sample facts carries over. Use only your own, authorized non-sensitive material. Expect the same kind of complete result using your inputs. Check every column against the owner’s rules. Apply the dictionary yourself to a valid row, a blank-value row and an invalid row; check the full key, not just the first field. Ask a colleague to try only if appropriate; do not claim that review has already happened. If a detail differs, paste: “This part is wrong: [quote it]. My input says: [quote the exact source line or rule]. Return a complete corrected replacement and explain the change.” Recheck it yourself. If the source is missing or contradictory, leave that part unresolved and obtain the needed fact before using it.
Your starting prompt
Document my small table as a data dictionary. Use only the owner-supplied definitions to state what one row represents and give each column’s meaning, type, unit, allowed values or range, missing-value meaning and example. Explain the unique-key rule and any cross-column conditions. Distinguish known rules from questions for the owner; never infer that a blank is zero or that an identifier’s apparent pattern is mandatory. Apply the dictionary to my supplied ordinary and deliberately invalid test rows without modifying them.
MY INPUT
Exact column names and complete non-sensitive example rows: [replace with complete non-sensitive details, or state unknown]
Owner-supplied meanings and what one row represents: [replace with complete non-sensitive details, or state unknown]
Types, units, allowed values and range rules: [replace with complete non-sensitive details, or state unknown]
Meaning of blanks, zero and special markers: [replace with complete non-sensitive details, or state unknown]
Unique-key and cross-column rules: [replace with complete non-sensitive details, or state unknown]
Separately labeled test rows to accept, reject or hold: [replace with complete non-sensitive details, or state unknown]
Use only MY INPUT for facts, names, dates, counts and constraints. Ask about essential missing information or conflicting rules instead of guessing. Do not import any practice-example values. Return the complete requested result, unresolved questions and checks tied to my input. Do not send, publish, collect responses, buy, book or change another service.
Optional: improve the result
Test dates and duplicate keys
Exercise the dictionary’s boundaries on separate fictional rows. This optional extension uses only the original synthetic sample. Return to the sample chat, or paste its complete Input into a new chat first. To use it with your own material, replace the source and every dependent rule, number and check explicitly.
Advanced prompt
Using only the original dictionary, evaluate three independent proposed additions: P01 | 2026-10-02 | 2 | sheets; P01 | 2026-10-01 | 0 | sheets; P03 | 2026-02-30 | 1 | sheets. For each, state accept or reject and the exact rule. Treat each as a separate addition to the unchanged original table, not a cumulative batch. No source edits.
Show, then try
See the shape of the result.
This input is synthetic. The output is a written illustration of the intended result, not a verified AI-tool run.
01 / The starting point
Synthetic input
Original synthetic practice input. The people, organizations, records and events below are fictional.
Complete fictional paper-count table. Vertical bars separate fields; the empty qty field below is intentionally blank.
item_id | counted_on | qty | unit
P01 | 2026-10-01 | 0 | sheets
P02 | 2026-10-01 | | sheets
Owner definitions:
- One row is one item’s count record for one calendar date; the record may say it was not counted.
- item_id is a nonempty text identifier. Preserve its characters and leading zeros. No further identifier pattern is supplied.
- counted_on is a required valid calendar date written YYYY-MM-DD; no clock time or time zone is stored.
- qty is either a nonnegative whole number or empty. Zero means counted and none observed. Empty means not counted. Negative numbers, fractions and text are invalid. No maximum is supplied.
- unit is required and must be the exact text sheets in this table.
- item_id plus counted_on together must be unique; the same item may occur on a different date. This two-field combination is called a composite key.
Also explain this separate deliberately invalid test row, which is not part of the two-row source: P03 | 2026-10-01 | -1 | sheets. Do not clean or change the source.
Copies text only. For image or file tasks, use the source files named in the steps too.
Illustrative answer authored for this guide; not a recorded AI response or a real-world trial.
What one row means
One item’s count record on one date. A row can record that the item was not counted; a present row does not guarantee a numeric observation.
Data dictionary
item_id
- Meaning: the item being recorded. Type: nonempty text. Measurement unit: none.
- Validity: preserve the exact identifier, including leading zeros; no additional pattern is established.
- Missing rule: blank is invalid. Example: P01.
counted_on
- Meaning: the calendar date of this item’s count record. Type: valid date in YYYY-MM-DD form. Measurement unit: calendar date, not a duration.
- Validity: required real calendar date; a date-shaped string alone is insufficient.
- Missing rule: blank is invalid. Example: 2026-10-01. No time or time zone is represented.
qty
- Meaning: observed number of sheets, or no observation yet. Type: nonnegative whole number or empty. Unit: sheets, as specified by unit.
- Validity: 0, 1, 2 and other nonnegative whole numbers are allowed; no maximum is supplied. Fractions, negative values and text fail.
- Missing rule: empty means not counted; 0 means counted with none observed. Example: P01 has 0; P02 has an empty qty.
unit
- Meaning: what qty counts. Type: required text label. Allowed value: exactly sheets.
- Missing rule: blank is invalid. Example: sheets. Do not treat packs as an equivalent value.
Whole-table identity rule
The pair item_id + counted_on must be unique. P01 on another date would be a different valid key. Repeating P01 on 2026-10-01 would violate the key even if its quantity matched.
Apply the definitions
- P01 / 2026-10-01 / 0 / sheets: valid; zero sheets were observed.
- P02 / 2026-10-01 / empty / sheets: valid record; quantity is unknown because it was not counted. It cannot be included as a zero in a complete total.
- Separate test P03 / 2026-10-01 / -1 / sheets: invalid qty because -1 is negative. Do not silently repair it to 0.
Owner questions if needed later
Is there an identifier pattern or quantity maximum? Neither is supplied, so this dictionary does not impose one. The two-row source remains unchanged.
These checks use the worked example and its particular facts. With your own material, check your actual names, numbers, sources and rules; do not reuse the sample’s values as the answer.
Unsaved checks stay in this tab and reset when you reload. If you choose Save guide, its current example checkmarks are kept in this browser when storage is available.
!
A few things to watch for.
A list of pretty headings is not a dictionary: it must explain how values are interpreted and what makes a record invalid.
A guessed identifier format can reject genuine records. Record unknown constraints as questions, not rules.