Many companies that have reached a mature observability practice continue to rely on auto-instrumentation, often without realizing it’s quietly draining their budgets. At this stage, observability engineers start spending countless hours chasing down cost anomalies, unaware that the problem may lie not in how much they collect, but what they collect.
What if the key to a more efficient telemetry pipeline lies inside the traces themselves?
In this talk, we will show how manual instrumentation can deliver leaner and more business-oriented telemetry. By comparing traces produced through automatic and manual instrumentation, it will be demonstrated how controlling spans and their attributes can cut data volume by up to 60%, reducing resource overhead and storage costs while improving signal-to-noise ratio.
Attendees will learn the do’s and don’ts of span and attribute management, how to identify unnecessary telemetry metadata, and practical steps to achieve higher observability efficiency.