Global distributed DB: balancing strong vs eventual consistency

[Lily.T]

I'm a lead engineer on a team developing a new distributed database system, and we're facing a critical design decision regarding our consistency model. Our system is intended for a global-scale application where data is replicated across multiple regions to ensure low latency for users worldwide. The core tension is between offering strong consistency guarantees, which simplify application logic but can increase latency for cross-region writes, and opting for eventual consistency, which improves performance and availability but places a significant burden on application developers to handle potential conflicts and stale reads. We've built prototypes for both approaches, and the performance benchmarks under simulated network partitions are starkly different. The business stakeholders are pushing hard for the lowest possible latency to compete with existing solutions, but our early adopter developers are expressing serious concerns about the complexity of building on an eventually consistent foundation. I'm responsible for recommending the final architectural direction. For engineers who have designed or worked extensively with globally distributed data stores, what were the key factors in your choice of consistency model? How did you quantify the trade-off between developer experience and operational performance, and were there any hybrid approaches or tunable consistency levels that you found provided a good practical compromise? Furthermore, how did you communicate the implications of this foundational choice to non-technical stakeholders to secure buy-in?