Ask HN: How do you reduce observability data volume without killing useful ones?

Different teams and services have very different needs. What’s critical for one team might be noise for another. Being too aggressive and using blunt, one-size-fits-all policies can hurt debugging or alerting.

Has anyone found flexible and efficient approaches that worked? Would love to hear how others balanced cost vs value without over-simplifying. Open to tools, strategies, or lessons learned.

Thanks!