# Insights

A computed view over one time window: how much came in, how much of it was errors, when the peaks were, and which issues account for the most volume.

## What is measured

- Total events and error events for the window, and the error rate between them.
- **Peak events per minute**: the busiest single minute in the window.
- **Active minutes**: how much of the window actually saw traffic.
- **Level totals**: the split across severity levels.
- **Open issues**: how many remain unresolved.

## Top issues

The issues producing the most events in the window, each with its culprit, level, status, event count and when it was last seen. This is the actionable part: a small number of issues normally account for most of the volume, and fixing those moves the error rate more than working down the list in order.

## Reading the timeline

Events are bucketed per minute and split by level, so a spike shows both its size and its composition. A jump made of warnings reads very differently from the same jump made of errors.

Peak per minute and active minutes are the pair worth watching together. A high peak across few active minutes is a burst, usually tied to a deploy or a batch job, while a steady rate across the whole window is a systemic problem.

## Empty results

Everything here is scoped to the selected window and environment. A new project, a window shorter than your reporting interval, or an environment filter matching nothing will legitimately render empty rather than erroring.