Overview
The Trimming node is a data preparation node in a Weave workflow that is responsible for filtering out the data on quantiles. Configuration consists of column selection (which columns are available), an active trimming column, a percentile value (0 to 1, the fraction of rows to trim), and a direction (First trims from the smallest-value end; Last trims from the largest-value end). Trimming is the standard Weave operation for removing a fixed percentage of records from the extremes of a distribution.
Trimming nodes are added from the Data Preparation category of the node panel. Configuration has more parts than typical Data Preparation nodes — column selection (Select Fields), active column choice (dropdown above Select Fields), percentile value, and direction. The percentile value uses the same 0-to-1 quantile scale as Outlier Treatment’s Min/Max bounds.
Note: “Weave” is the current name for what earlier interface versions called “Workflow.” This reference uses “Weave.”
When to use it
- Removing a fixed percentage of rows from the extremes of a numeric distribution — common pre-processing for statistical models sensitive to outliers.
- Stripping the highest values when high-end outliers represent special cases (e.g., enterprise-tier customers in a small-business analysis).
- Quartile-based row removal — trim 25% from one end to drop a full quartile.
- Building robustness — combining Trimming with downstream statistical operations to reduce the impact of extreme values.
Configuration
Column selection (Select Fields)
| Select (button) | Opens the Trimming column-selection popup. |
| Select All | Checks every column in the popup. |
| Individual column checkboxes | Pick specific columns. The popup determines which columns are available for the active-column dropdown. |
| Apply | Commits the column selection. |
Select N Percentile
| Value range | Between 0 and 1 inclusive. |
| Validation | “Enter valid percentile (0-1)!” appears when an out-of-range value is entered. |
Select direction to trim data
| First | Trims the configured percentile of rows from the top of the sorted column (the rows with the smallest values). Use to remove low-end outliers or small values. |
| Last | Trims from the bottom of the sorted column (the rows with the largest values). Use to remove high-end outliers. |
Key behaviors
Trimming is directional and percentage-based. Unlike Outlier Treatment’s IQR Method which drops both tails of the distribution based on quantile boundaries, Trimming drops a fixed percentage from one specific end (First or Last). The Designer explicitly chooses what fraction to drop and from which direction. This makes Trimming more predictable in row-count reduction — 0.10 always reduces row count by exactly 10% — but less adaptive to the actual data distribution.
Sort by the active column, then trim. The active trimming column is the sort key. Trimming sorts the data by this column’s values and removes rows from the chosen direction. The other columns in the row are dropped along with the row even though they didn’t participate in the sort — Trimming removes entire rows, not values.
Percentile is between 0 and 1, not 0 to 100. The 0-to-1 quantile scale is shared with Outlier Treatment’s Min/Max bounds. Enter 0.30 to trim 30%, not 30. The UI validates this with the “Enter valid percentile (0-1)!” error for out-of-range values.
Trimming preserves column values. Unlike Outlier Treatment’s Z-score Normalization (which transforms values) or Winsorization (which clamps values), Trimming only removes rows. The remaining rows have their original column values unchanged. This makes Trimming the right tool when extreme rows should be excluded but the rest of the data must remain in its original form.