Alex The Analyst on YouTube · Open on YouTube
Professional practice: how analytics teams actually work · Competent · Lesson 31 of 99 · 51 min video, then about 6 hours of practice
Cleaning data at scale, reproducibly
After this you can profile a large table, find duplicates, inconsistent categories and impossible values, and fix them in code that can be rerun.
Why it matters
At work, cleaning is not a one-off; the same source arrives again tomorrow. Clean in SQL or pandas, never by hand, and make every rule explicit and countable (how many rows did this affect?). Profile before cleaning: row counts, distinct counts, NULL rates, min and max per column. The traps are deleting data silently, and cleaning so aggressively that you remove the very anomalies the business needs to know about.
How to use this resource
A full data-cleaning project in SQL: deduplication with window functions, standardising text, handling NULLs and dropping columns safely.