Overview
The Imputation node is a data preparation node in a Weave Workflow that fills null (missing) values in selected columns using a configured Fill With method — Mean, Median, Mode, or a user-specified value. It sits between source nodes (Data Import) and terminal nodes (Result, Output), or between other transformation nodes, replacing nulls with computed or supplied values before the data reaches downstream operations that can’t tolerate missing data.
Imputation nodes are added from the Data Preparation category of the node panel. Configuration is two steps: select columns via the Select button (opens a column-selection popup with a Select All option), then pick the Fill With method via mutually exclusive radio buttons in the right panel. Only one Fill With method applies per Imputation node.
Note: “Workflow” is the in-UI term for what some Weave documentation calls a pipeline. This reference uses “Workflow.”
When to use it
- Filling nulls before joins, where unmatched rows would propagate the missing data into downstream calculations.
- Preparing data for analytical models or statistical operations that don’t accept null values.
- Replacing missing measurements with a representative central value (Mean for symmetric distributions, Median when outliers skew the mean).
- Substituting a known default (User Specified Check) for nulls — e.g., “Unknown”, 0, or a sentinel value documented elsewhere in the data dictionary.
Fill With methods
Imputation supports four mutually exclusive fill methods. Pick exactly one per Imputation node — they cannot be combined within a single node.
| Mean | Fills nulls with the column’s arithmetic mean (sum of non-null values ÷ count of non-null values). Numeric columns only. | Numeric columns with roughly symmetric distributions and no extreme outliers. |
| Median | Fills nulls with the column’s median (middle value of the sorted non-null values). Numeric columns only. Robust to outliers. | Numeric columns with skewed distributions or extreme outliers where Mean would be misleading. |
| Mode | Fills nulls with the column’s most frequent value. Works for numeric and categorical columns. | Categorical columns, or numeric columns where the most common value is more representative than the central tendency. |
| User Specified Check | Fills nulls with a user-supplied value. | When the appropriate fill value is known and documented (e.g., 0) and shouldn’t depend on the column’s existing data. |
Column selection
Columns to impute are picked via a popup with a Select All shortcut and individual column checkboxes.
| 1. Open the Select dialog | Click the Select button under Select Fields in the right panel. An Imputation popup appears with the upstream columns listed. |
| 2. Choose columns | Use Select All to check every column, or check columns individually. Columns left unchecked pass through unchanged — their nulls remain. |
| 3. Apply | Click Apply to commit the column selection. The selected columns are now scoped to the Fill With method picked in the right panel. |
Key behaviors
Imputation is a transformation node. It sits in the middle of a Workflow between source and terminal nodes. Imputation has one input and one output; it transforms data passing through without affecting the source Datasource.
Two-step configuration: columns, then method. The Select button popup chooses which columns the Fill With method applies to. The right-panel radio buttons choose which method to apply. Both steps are required — picking a method without selecting columns produces no imputation; selecting columns without a method does nothing either.
Fill With methods are mutually exclusive. Unlike Cleansing’s stackable checkbox operations, Imputation’s Fill With methods are radio buttons — exactly one applies per Imputation node. To use different methods for different columns (e.g., Mean for numeric, Mode for categorical), chain multiple Imputation nodes.
Non-null values are unchanged. Imputation modifies null cells only. Cells with existing values pass through unaffected. This makes Imputation safe to apply broadly — including non-null-containing columns in the selection doesn’t risk corrupting their data, the operation just has nothing to do for those cells.
Use Imputation early in the chain. Imputation works best when applied before joins, aggregations, or other operations that propagate or amplify the impact of missing data. A Join with nulls on the join column drops rows; an Aggregate over nulls produces nulls in the result. Imputing first removes those concerns.