Observability solutions often require deploying agents or collectors to manage telemetry data such as logs, metrics, traces, or profiles. While directly sending this data to storage systems might seem simpler, it poses challenges like vendor lock-in, limited data enrichment, and issues with redeployment and system downtime. Alternatively, a pull-based method, exemplified by Prometheus, allows for backend flexibility but is tailored mainly to metrics and may not suit short-lived applications. A hybrid approach, involving a collector or agent, offers a balance by decoupling applications from storage backends, enabling seamless transitions, data enrichment, and simplified authentication management. The OpenTelemetry Collector is highlighted as a versatile tool for this purpose, with customizable configurations that allow data to be received, processed, and exported efficiently. Deployment strategies for collectors vary, including single-instance, sidecar, and node collectors, each with scalability considerations. Effective implementation requires separating telemetry types, chaining collectors for layered processing, maintaining protocol consistency, and prioritizing instrumentation from the application's inception.