Why Averages on Dashboards Can Mask Critical Performance Problems for Key Users
A software engineer at DEV Community, Sergey Shinder, explains how a 120ms average response time on a latency dashboard concealed serious slowdowns affecting a subset of users. While the metric satisfied leadership, customer support was fielding complaints about an unbearably slow app — both sides were correct, but the average obscured the gap. Shinder found that switching to percentile-based metrics, particularly the 99th percentile, revealed that while most requests completed in 90ms, the slowest experiences stretched to 6 seconds. The real insight came from slicing data by endpoint, customer tier, region, and cache state, which pinpointed the problem to a specific query pattern affecting high-volume customers. He draws a distinction between monitoring — dashboards built around known questions — and observability, which allows teams to investigate new questions mid-incident without deploying new code.
This is an AI-generated summary. ShortSingh links to the original source for the complete article.
Discussion (0)
Log in to join the discussion and vote.
Log in