MSTR's underlying technology is best in class. The answer to "is [X] possible?" is always "yes". The dependency chain is unbreakable, which is really comforting from an architecture development standpoint. One of the pain points we had with Looker prior to switching to MSTR was that it didn't have a robust dependency chain (we did like using Git for version control though).
The SQL generation is also *very* readable, in contrast with many of MSTR's competitors.
The in-memory cubes are powerful - 2B row limit per CPU. We used these cubes to build some otherwise prohibitively large dynamic dashboards.
UI, support, stability. Support is predicated on a vast knowledge of MSTR's intricacy's, including deep logging features which do not feel Cloud-worthy. Cloud is clearly a new philosophy for them. Professional services is hit or miss.
The "managed service" aspect of the Cloud product still relies on a deep understanding of the platform and its nuances and jargon.
If starting from scratch, be sure to include nimble, scalable data sources such as Snowflake or BigQuery
Self-service access to complex data that would be otherwise inscrutable to business teams. We've found numerous asymmetrical opportunities to improve our business.
We've also productionized large chunks of our client-facing work - the dashboards help us find the fires and put them out, find the smoke where there might be more fires, and find the successes we might have otherwise missed.