Data Cleaning Exercises

Module 20 · Job-ready delivery · 8 min

Write the handoff note

You'll be able to: Hand over a cleaned dataset with the rules, evidence and open questions someone else needs.

Before this: Hand off a cleaning job: rules, exceptions, reconciliation

The idea

A handoff note lets a colleague trust, rerun and question your work without asking you.

Include: the rules in the order applied, assumptions, a reconciliation table, exceptions by reason with examples, open questions with owners, and how to rerun.

Keep it short and factual; link to the code and the exception file rather than pasting them.

Worked example

Input: orders.csv → clean_orders + exceptions

Visual
(document) Rules · Reconciliation · Exceptions · Questions · Rerun
SQL (DuckDB)
SELECT reason, COUNT(*) FROM exceptions GROUP BY ALL ORDER BY 2 DESC
pandas
exceptions["reason"].value_counts()

Result: Input 1,000 = clean 951 + exceptions 49 (bad_timestamp 21, unknown_customer 18, duplicate 10)

Common mistakes

  • No counts.
  • Assumptions left implicit.
  • No way to rerun.

Check yourself

What proves nothing was lost?

The reconciliation: input rows equal clean rows plus exceptions.

Practice it

Clean the export and report exceptions by reason.

Next: Clean an order-line export