I have a general question about Powerpivot usage.
A friend of me asked to support him in a small Powerpivot project:
All the Powerpivot projects I previously did were pretty easy (create SQL Views, connect them, build relationships, create nice charts/tables and some extra measures in Excel).
In this project I'm not very sure about the best approach:
In the moment I really don't like the current approach for the following reasons:
Better ideas?
A friend of me asked to support him in a small Powerpivot project:
- He has to create a report every few months. This report is created in Excel but with many editorial stuff: Summaries, formatting, and additional categories which are not stored in the database. The additional categories are based on actual database data but reordered and recalculed with additional formulas.
- In the moment he has to create this report every 3-6 months. Every time he has to work for 3-5 days to finish the report.
- The database contains about ~2 million data rows in several tables.
- He thinks Powerpivot is a strong tool to access and aggregate/calculate this database data.
- In the end he wants his report to be updated in a few minutes after refreshing the database data.
All the Powerpivot projects I previously did were pretty easy (create SQL Views, connect them, build relationships, create nice charts/tables and some extra measures in Excel).
In this project I'm not very sure about the best approach:
- The data in the database has to be recalculated/aggregated with many different and complex formulas. In total I expect about 30-40 formulas. They usually look like: If payergroup=x and productgroup=y then summarize all results except subproductgroup y1. Every group has it's own formulas. This rules can be stored in the database (as an extra view) or applied via measures. The artificial categories are company units, sections or projects which are not stored in the database. Unfortunatly many projects intersect so the data has to be restructured with complex rules.
- My current idea is to create several Powerpivot tables based on the original database tables and use measures for the complex calculations.
- Unfortunatly the resulting Powerpivot tables don't match his requirements. He wants to summerize them in an nice report which is based on artificial categories which are not stored in the database.
- The easiest solution seems to keep his current report structure and use Excel formulas to reference the Powerpivot tables stored in an addition Excel sheet. He might use the vlookup function in Excel to regroup existing categories in his report specific categories, so his Excel formulas won't rely on the actual column index in the pivottable.
In the moment I really don't like the current approach for the following reasons:
- It will result in many Powerpivot tables, measures and individual formulas
- It doesn't look very solid in general
Better ideas?