Operational Burden Assessment
Quantifying the ongoing engineering effort a self-managed component requires, so it can be compared honestly with a managed service's price.
Managed services look expensive because their cost is a visible line item, while self-managed cost is distributed across salaries and therefore invisible. The comparison is routinely made on the wrong basis.
What self-managing actually includes, and should be estimated explicitly: capacity planning and scaling, patching and version upgrades (including the major-version migrations nobody plans for), backup and verified restore, high availability and failover testing, monitoring and alerting, on-call burden, security hardening and compliance evidence, and the expertise to run it well — including what happens when the person who has it leaves.
Expressed in engineer-days per year, most self-managed data stores cost considerably more than the managed equivalent.
The genuine reasons to self-manage remain: scale at which the managed premium becomes very large; a feature or version the managed service does not offer; regulatory constraints on where and how it runs; and existing deep expertise that is already funded.
The reason that is not good enough on its own is control. Control has a price, and the assessment is what makes that price visible so it can be a decision rather than a preference.
Note the reciprocal cost too: managed services impose maintenance windows and forced upgrades on someone else's schedule, which must be designed for.