There are a number of things that can cause poor Tableau dashboard performance. Therefore there are a number of things to check that could improve your Tableau dashboard performance.
Check the Data
1. Unless your data source is a super fast analytics optimised database, such as a columnar database, then use a Tableau Hyper file. A Hyper file is Tableau’s own super fast database, optimised for analytics. It can aggregate and return millions of rows in seconds. In addition it offers multiple other benefits. However this article isn’t about the benefits of Tableau extracts. Fortunately Tableau have written all about the benefits in their paper on the Tableau Data Engine.
2. Optimise your extract. It’s in the toolbar under Data – Extract – Optimize. If your extract is pulling from a SQL Server and the extract refresh speed isn’t good there are things you can do to speed up the refresh of Tableau extracts based on SQL.
3. If connecting live your dashboard performance depends on the performance of your underlying database. In addition each component on your dashboard means a separate query to run against the database. Therefore the more objects on the dashboard, the slower it will be.
4. Create calculated fields in your data source BEFORE publishing the data source to the server. This optimises the calculation. Creating calculated fields in the workbook slows things down. Note, some calculations can’t be optimise / pre-aggregate, such as COUNTD, therefore they will always be slower.
Changes within the workbook
5. Use Window_sum instead of Total. This post gives details on that and also advises how to speed up the Window_xxx functions.
6. With bins it’s quicker to use the Tableau auto-created bins (right click on a Dimension and select Create Bins) than to create custom bins using a calculated field. If the bins need to be custom, if possible create the bins in the source data rather than in Tableau. If creating bins by calculated field then refer to #4 above.
7. Quick filters can be slow – the more you have on the dashboard the longer it will take to render, especially if they’re set to only show Only Relevant Values. Each time a filter is changed the quick filters recalculate when showing relevant values.
8. FIXED Level Of Detail calculations are slow against large data sets.
If there are more suggestions please send them my way and I’ll update this article accordingly.