Transformation and modelling.
WHAT THE TRANSFORMATION STAGE DOES
Cleans and reshapes data before it enters the model.
WHAT TO DO THERE
Remove columns you do not need Correct data types Handle missing values deliberately Split and combine columns Filter to relevant rows
WHY REMOVE COLUMNS EARLY
Every retained column costs memory and refresh time.
WHAT TO PUSH BACK TO THE SOURCE
Filtering and aggregation, where the source can do it.
WHY
It is faster there, and less data is transferred.
WHAT A DATE TABLE IS
A dedicated table of dates, marked as such.
WHY IT IS ESSENTIAL
Time-based analysis depends on it, and calculations behave incorrectly without one.
WHAT TO DEFINE
Measures, rather than calculated columns, where possible.
WHY
Measures are calculated at query time and consume no storage.
WHAT TO NAME CLEARLY
Every table, column and measure.
WHY
Report builders and readers see these names.
WHAT TO HIDE
Technical columns not intended for report use.
WHAT TO DOCUMENT
What each measure means, in business terms.
