A little about Tableau Reporting

Design Workbook:

1. Keep it simple.
2. Use only the data you need.
3. Use filters efficiently (use context filter if possible).
4. Not all mark types are equals.
5. Use Fixed dashboard size for the fastest loading. Automatic size is not recommended because Tableau will generate every time it is turned on.

Is it my calculation?

1. Data physics -> type data: number/bool > date >> string; CNTD most expensive.
2. Use right calculation -> LOD paling pelan; R/phyton can be slow because must wait data serialized from/to R.

Fix my query design

1. Having your data all in one place is faster than pulling it from multiple sources.
2. the less you have to manipulate your data, the faster it will be -> union, pivot, calculation, etc.
3. Trust Tableau to do its job -> avoid customSQL and stores procedures.
4. Tableau does lots of stuff to make your workbooks fast -> parallel execute.

Fix my data source design

1. Extract are an easy way to make things go fast -> hide unused fields.
2. Tune your database for optimal performance.
3. Use data server for governance -> have data expert/stewards optimize the connections and share it with business users; use extracts across multiple workbooks/share extracts as possible.
4. Use native drivers where available -> better than ODBC/JDBC
5. Test query performance on the server.
6. Use cold (raw) / warm (prepared) / hot (aggregated) strategies.

Re-design bad environment

1. Upgrade -> physical server faster; Tableau need RAM than CPU.
2. Be prepared to test on the server.
3. Keep interactive users and extract refreshes separated.
4. Monitor and tune.
5. Virtual vs physical -> Tableau need dedicated resources for the best performance, physical better than VM.

https://www.tableau.com/learn/whitepapers/designing-efficient-workbooks