Purpose
Use this lesson to turn messy data symptoms into explicit, repeatable transformations. The aim is to produce work that a colleague can review and act on, not merely text that sounds convincing.
Lesson Explanation
Scenario: A customer file has duplicate emails and inconsistent country names.
Better instruction: Propose cleaning rules: trim spaces, standardize country names to an approved list, flag duplicate emails, and preserve the original value in a separate column.
Data cleaning must be reversible and auditable. Define source columns, transformation rules, approved values, and what to flag rather than overwrite. Ask the AI for rules and test cases before applying changes to production data. Original values remain available, duplicates are flagged, and countries follow an approved list.
Work in a short cycle: provide the relevant input, state the result and limits, inspect the draft against the acceptance criteria, then revise the instruction when a requirement is missing. Keep the final human decision with the person responsible for the work.
Practice Questions
-
A prompt asks an AI to “clean up this customer list” with no further detail. What does this lesson say data cleaning must be?
View Answer
Reversible and auditable – the original values should be preserved, not simply overwritten.
-
What does this lesson recommend doing with the original value when a data field is corrected or standardized?
View Answer
Preserving it in a separate column, rather than overwriting it.
-
Write a prompt for cleaning a list of country names with inconsistent spelling, following this lesson’s guidance.
View Answer
Something like: “Propose cleaning rules: standardize country names to this approved list, flag any that don’t match, and preserve the original value in a separate column.”
-
A data-cleaning prompt doesn’t specify what counts as a duplicate or how to handle one. What does this lesson recommend defining?
View Answer
Rules for what to flag (such as duplicate emails), alongside transformation rules and approved values.
-
What does this lesson recommend requesting before applying any cleaning changes to production data?
View Answer
Rules and test cases from the AI, reviewed before applying changes to actual production data.
-
A cleaning process overwrites the original spelling of a customer’s address without keeping any record of the original value. What does this lesson say about this approach?
View Answer
It violates the principle that cleaning must be reversible and auditable – the original value should have been preserved in a separate column.
-
A prompt for cleaning a product catalog specifies source columns, transformation rules (standardizing units), and approved unit values. What is one more element this lesson recommends adding?
View Answer
What to flag rather than overwrite – i.e., handling of ambiguous or unexpected cases.
-
Why does this lesson emphasize testing cleaning rules before applying them to production data?
View Answer
To catch problems with the proposed rules on a smaller scale before those rules affect the actual, live dataset.
-
A data-cleaning prompt flags duplicate customer emails for review, rather than automatically deleting them. What principle does this reflect?
View Answer
Reversibility and auditability – flagging preserves the ability to review and decide, rather than making an irreversible automatic change.
-
A prompt asks the AI to “standardize this data” without naming an approved list of acceptable values. What is missing, per this lesson?
View Answer
Approved values that define what standardized data should look like.
-
A cleaning workflow applies proposed rules directly to the live customer database without first testing on a sample. What does this lesson recommend instead?
View Answer
Testing rules and test cases before applying changes to production data.
-
A completed data-cleaning specification names source columns, transformation rules, approved values, what to flag, and preserves original values in a separate column. What does this reflect?
View Answer
A complete, well-structured cleaning specification consistent with this lesson’s full guidance.