Data Partitioning Strategies for Multi-Tenant SaaS: Balancing Isolation and Cost?

Hey everyone, I'm working on the design for a new multi-tenant SaaS platform and we're debating data partitioning strategies. We've considered separate databases per tenant, shared databases with schema-based partitioning, and shared databases with row-level partitioning. Each has its pros and cons regarding data isolation, backup/restore complexity, and operational cost. For those who've built similar systems, what strategies have you found most effective, especially when dealing with a wide range of tenant sizes and varying compliance requirements? Are there any less common approaches worth exploring, or specific tooling that makes one strategy significantly easier to manage in production? I'm particularly interested in hearing about experiences with horizontal scaling and how different partitioning schemes impact it.