What are the common causes of wrong numbers in a workflow?

Q: What are the common causes of wrong numbers in a workflow?

Almost always a row count changing somewhere unnoticed. A Joiner with duplicate keys on the right multiplies rows and inflates every downstream measure. A Row Filter excluding more than intended silently removes data. A GroupBy drops columns that were neither grouped nor aggregated. A Column Appender misaligns tables that were not genuinely parallel. None of these errors — the workflow runs and the numbers look plausible. The defence is checking the row count after every node that can change it,

Question

Accepted Answer

Almost always a row count changing somewhere unnoticed.

A Joiner with duplicate keys on the right multiplies rows and inflates every downstream measure. A Row Filter excluding more than intended silently removes data. A GroupBy drops columns that were neither grouped nor aggregated. A Column Appender misaligns tables that were not genuinely parallel.

None of these errors — the workflow runs and the numbers look plausible.

The defence is checking the row count after every node that can change it, rather than only at the end. Finding it at the end means bisecting the whole workflow.