It is easy to demonstrate a neat transformation on clean data. The more useful test includes missing values, inconsistent dates and rows that only look like duplicates.
What would make you trust the preview? Would you keep a record of each transformation so the process can be repeated?
.
