Inspecting Workflow

When you're inspecting a new dataset, here's what "good data" looks like:

1. Reasonable null counts – A few missing values are normal, but if 50% of a column is null, that's usually a problem.
2. Expected data types – Numbers should use an appropriate numeric type, not object (i.e., text).
3. Realistic min/max values – Negative ages? Dates in the year 2200? Investigate.
4. Consistent column names – If you see user_id and userId, something's off.
5. Expected row counts – Does the number of rows match what you were told about the data?

Real-world data is often messy, so you need to know what you're dealing with. If you see any of the issues mentioned above, it's a sign that you need to do some data cleaning before you can conduct any meaningful analysis.