Forum Discussion
Normalization
Hello everybody,
seems my previous post got lost. I have a question in regards to if I should go through an additional normalization step during the import proces to improve system performance.
The fact table has about 100k (max. 300k one day) rows and 12 columns. Values to calculate with are only in one column. This table represents the financial reports (P&L, BS) for 3 years incl. actual, Budget and FC on a monthly level and for all items a balance sheet and P&L has to offer. 6 of those columns are representing organizational levels (Legal entity, plant and profit center) and show for each a key (integer) and name (text).
Would it be better to delete the columns with the names and have them in dim-tables, or keep them in the fact table?
Many thanks for your opinion in advance.
Daniel
danielboi , It always better to have a dimension table and be in a Star schema
refer
https://www.youtube.com/watch?v=vZndrBBPiQc&feature=youtu.be
https://www.sqlbi.com/articles/the-importance-of-star-schemas-in-power-bi/
4 Replies
danielboi , It always better to have a dimension table and be in a Star schema
refer
https://www.youtube.com/watch?v=vZndrBBPiQc&feature=youtu.be
https://www.sqlbi.com/articles/the-importance-of-star-schemas-in-power-bi/
Thank you Amit. I read that "over-normalizing" could be detriment to read performance. That was triggering my question. 🙂
danielboi , I think you can work very well with a single table. But when you need to use all and all selected then, in that case, having a dimension table help. I can use it all in one dimension without disturbing others. There many advantages of Star schema, once complex needs start coming, you will have most of the solutions around it.
