Microservices in distributed systems can produce huge amounts of telemetry data. But not all of this information is equally valuable when analyzing the broader system. Picking only data points that describe your system well is key for understanding your system while staying cost effective. Based on a real world example, this session discusses different sampling techniques such as head based and tail based sampling, and helps you find a suitable strategy to match your telemetry data. We will guide you through incrementally building architectures that evolve with your operation, empowering your system to adapt seamlessly. After this session, you will be able to select the most valuable data points from your telemetry data and design an architecture around it.