While the benefits of observability are clear, it requires expertise in many areas, has a high learning curve, and is technically complex and challenging. Even with observability, it is difficult to get meaningful results as expected because there is a lot of noise. You need to understand your signals and use anomaly detection and AIOps to reduce noise and improve accuracy.