Many of the metrics you would normally track outside of the product we collect inside of the product so it is also useful to our self hosted users. This includes cohort analytics, conversational development analytics, and cycle time analytics.
We want a single data warehouse and a central data model. We bring all relevant data to single storage place. Make it platform agnostic, so for example do not build it into Salesforce. We want to build data model to be used consistently across tools and teams. For example something as simple as unique customer ID, product or feature names/codes.
On the other hand we're open to pragmatic solutions linking for example Salesforce and Zendesk, if there are boring solutions available we'll adopt them instead of creating our own.
We want the tools to be open source so we can ship this as a product. Also see the "Complete BizOps" Google Doc. We'll use Singer, PostgreSQL, and Superset.
Warehouse: PostgeSQL, maybe later with a column extension. If people need SaaS BigQuery is nice option and Druid seems like a good pure column oriented database.
Display/analytics: Superset (Apache open source, mostchangedrepository) instead of the open source alternative Metabase which is Clojure based and runs on the JVM or Redash. Proprietary alternates are Looker and Tableau.