Read each requirement as if you were handing the output to another analyst. Some tasks ask for a calculated table, while others ask you to choose the safe sequence or identify a limit in the source. Keep those requirements separate: a correct total cannot repair an incorrect population.
This final set reviews output grain, functions, null behavior, refresh and loading. It uses fresh original prompts rather than repeating the timed mock questions. You can take as long as you need. After checking a task, trace the explanation against a few source rows and consider how the same steps would respond to an empty file or a missing lookup key. The hidden data checks repeatable transformations for the practical tasks.