A well-designed analytical model with ample resources can still struggle to produce adequate performance due to an inefficient visualization layer. The following list of techniques can be applied to Power BI reports and dashboards to reduce the query workload and avoid slower resource-intensive queries:  

• Create dashboards on top of reports to leverage cached query results representing the latest data refresh:
  
  • Unlike dashboards, report queries are sent and executed on the fly when Power BI reports are loaded.
  • Multiple dashboards can be linked together as described in Chapter 7, Designing Power BI Dashboards and Architectures.
  • If the dataset uses a DirectQuery or Live connection, take advantage of scheduled cache refresh as described in the Dashboard cache refresh section of Chapter 9, Managing the On-Premises Data Gateway.
• Avoid report visuals that return large amounts of data such as tables with thousands of rows and many columns:
  • Report visuals that require scrolling or which represent a data extract format should be filtered and summarized.
  • Report visuals that return more data points than necessary to address their business question can be modified to a lower granularity. For example, a dense scatter chart of individual products could be modified to use the less granular product subcategories column.  
• Ensure that filters are being applied to reports so that only the required data is returned:
  • Apply report level filters to only return the time periods needed (such as current year and last year).
  • Use visual level filters such as a top N filter as described in the Visual-level filtering section of Chapter 5, Creating and Formatting Power BI Reports.
• Limit the volume of visuals used on a given report page:
  • Optionally remove the interactions between visuals (cross-highlighting) to further reduce report queries.
• Understand which DAX measures are less performant and only use these measures when required:
  • For example, only use expensive measures in card visuals or within highly filtered visuals exposing only a few distinct numbers.