Finding them
The Statistics node reports missing counts per column. The Data Explorer node gives the same alongside distributions, which is usually the more useful of the two for a first look.
Missing values in KNIME are a distinct state — they are not empty strings or zero, and $col$ = "" does not find them. Use MISSING $col$ in a Rule Engine.
The Missing Value node
Set a strategy per column, or a default per data type:
| Strategy | Effect |
|---|---|
| Remove row | Drops any row with a missing value in that column |
| Fix value | A constant you supply |
| Mean / Median / Most frequent | Statistical fill |
| Previous / Next value | Carry forward or backward |
| Linear interpolation | For ordered numeric series |
| Moving average | Smoothed fill |
Each choice is a claim
Filling revenue with 0 asserts the sale was zero. Filling with the mean asserts it was typical. Those produce different totals and different conclusions.
Decide deliberately, per column, and record why.
Interpolation and carry-forward suit time series with genuine gaps, and are wrong almost everywhere else.
Removing rows
Setting Remove row on several columns can quietly discard most of the table, since it takes one missing value anywhere to drop a row.
Check the row count before and after — the node does not warn you.
Missing versus empty
A CSV can contain empty strings that look missing but are not. The String to Number or Cell Replacer nodes, or a Rule Engine converting "" to missing, normalise this — otherwise your missing-value strategy silently skips them.