Find the NaNs Hiding in Your Table

Real datasets are full of gaps. Before any analysis, you need to know where missing values hide, because they quietly skew every result you trust. 🔎

In pandas, an empty cell usually shows up as NaN, short for Not a Number. It is the standard marker for a value that simply is not there.

A common trap: treating NaN as if it were 0. Zero is a real measurement, while NaN means you have no measurement at all. They tell very different stories.

The method isna() returns a table of True and False, one flag per cell, marking exactly where values are missing across your whole DataFrame.

Chain isna().sum() to count missing cells in every column at once. True counts as 1, so each total is the number of gaps in that column.

Want one grand total of every gap in the table? Add another sum() to collapse the per-column counts into a single number.

Raw counts can mislead on big tables. Divide by len(df) to see the share of each column that is missing, which guides how seriously to react.

The opposite of isna is notna(), which flags cells that do hold a value. It is handy when you want to keep only complete rows.

A heatmap of isna() turns gaps into a picture, so clusters of missing rows or columns jump out far faster than scanning numbers.

To inspect problem rows, filter where any value is missing along each row. This shows you exactly which records need attention.

Mapping missingness before you fix anything keeps you honest. You decide how to handle gaps based on evidence, not on a guess about how bad they are.

You want the number of missing values in each column. Which expression gives it?

You now find missing data with isna, count it per column, view it as a percentage, and map it visually. Seeing the gaps is the first step to fixing them.