Blank cells are not zeros and they are not errors: they are unknown. Every statistic you take over a column with unknowns needs a stated rule for what to do with them. The default behaviours differ between languages, which is why the same dataset gives different numbers on each side of the team.
About 2 min of reading
Before this new lesson, one small question about the lesson you just finished. Recall it, do not look it up.
Average a column that has blanks, honestly.
Build a column with gaps
Before you run it, say out loud what it will print. Then open the answer.
Clean and count
Before you run it, say out loud what it will print. Then open the answer.
Write one function that turns a raw cell into a value or None, and use it in every analysis. Two different blank handling rules in one report is the bug that makes two sections disagree.
Before dropping rows with blanks, count them by group. Missing values that cluster in one region or one month are a signal about the collection process, not noise to be discarded.
int('') crashes; float('') crashes; treating '' as 0 invents data. Decide whether the row should be dropped, whether the column should be skipped, or whether the gap should be reported as a count of unknowns.
In Python, float('nan') != float('nan') and in JavaScript NaN !== NaN. So 'x == NaN' is always false; use math.isnan() or Number.isNaN() to test. This surprises anyone writing a filter for bad numbers.
'average 4.2 over 118 of 140 rows' is honest. 'average 4.2' hides that 22 rows had no value, which is often the interesting finding. Print the denominator with every statistic.
Each one is defined here on this site. Open the entry for the full definition and the other words it points at.