With a well-build datamodel could a dasboard run very smoothly with 50 M records even with quite small hardware-ressources and you could heavily slow down an application with a few k records if you makes (many) things wrong.
There is no best approach to build a datamodel and an application - it will be always depend on the kind of data, the number of user and the requirements of the business which data should be calculated and displayed in which way. Therefore you need to be a lot more specific with your question.
Check few points below
- Do you really need all the data? Check with user how many past years data they want? Last 5, 10 or N. This will reduce the file size, Records count, if file size is less QV performs well.
- Load only the required columns.
- Truncate Timestamp if you are not using it for analysis
This are the main points to consider while creating a data model.