Answering a question on Twitter:

While there is no dedicated transformation for selecting a random subset of rows (yet), it’s relatively simple to do with just a few transformations: calculate new column with a random number, then only keep rows where the random number is below some threshold (which can be defined as a parameter). This method is not very precise but can be sufficient.

In the example a sample dataset is generated. In a real-life case it can be substituted with an import transformation.

Sample project: Random subset.morph (2.8 KB)