Clear comparison. The point I would underline is that the tenancy model is rarely a pure technical choice. It usually follows from two business questions that are worth answering before any schema work: does any customer contractually need their data isolated, and how bad is it if one large tenant slows everyone else down.

For most early SaaS products we start with the pool model and row level security, because it keeps migrations and operations simple, and we design the tenant key so that moving a single large customer into its own schema or database later is a contained job rather than a rewrite. Writing that escape route into the original scope costs very little and removes a lot of fear when the first enterprise deal asks about isolation.

Did you find the bridge model worth its migration overhead in practice, or mostly a stepping stone?