Choosing Between ORM and Raw SQL for Data Access: Practical Considerations?

Hey everyone, I'm working on a new service and am at the point of deciding on the data access layer. I've used ORMs extensively in the past, and they offer quick development and often handle a lot of the boilerplate. However, I've also hit performance ceilings and felt limited when complex queries or very specific optimizations were needed, leading to dropping down to raw SQL anyway. Conversely, starting with raw SQL gives maximum control but can be more verbose and slower to initially develop, especially with schema changes. What are your thoughts on where the balance lies? Are there specific scenarios where you always default to one over the other? For instance, for simpler CRUD, ORM seems obvious, but what about reporting or highly-optimized read paths? Interested to hear about your experiences and any frameworks or patterns you use to manage this trade-off effectively.