# Metrics

Source: https://instatus.com/help/observability/metrics

Metrics are numbers your services report over time: request duration, memory, CPU, queue depth, anything you count. Instatus shows them biggest change first, so what moved recently is at the top.

Open **Observability** → **Metrics** on your status page.

![Metrics as cards, each with its current value, its change against the previous hour and a chart](https://instatus.com/help/observability/metrics.png)

Metrics arrive once your app sends them: tick **Metrics** when you [copy the setup prompt](https://instatus.com/help/observability/send-logs#option-1-copy-a-prompt), or use auto-instrumentation, which reports request duration, memory and CPU for you.

## Read the cards

Each card shows a metric’s current value and how much it changed against the previous period. With **Last 1 hour** selected, that’s the hour before. A red change is worth a look.

- **Sort** switches between **Biggest change** and **Name**.
- **Service** shows only the metrics of one service.
- The search box finds a metric by name, such as `http.server.duration`. Press **/** to jump to it.
- The time range sets both the chart and the period it is compared with.

## Open a metric

Click a card. The chart shows the metric over the time range, with one line per service, and a table below compares them: the value now, the average, the peak, and the change.

![A metric with one line per service, a highlighted spike and a link to the slow traces behind it](https://instatus.com/help/observability/metric-detail.png)

- **Show** picks how values are combined. Durations offer `p50`, `p90`, `p95`, `p99`, average and max. Gauges such as memory offer average, min and max. Counters offer rate per second and sum.
- **Split by** draws one line per **service**, or a single line for **none**.
- Point at a service in the table to highlight its line. Click it to keep it highlighted.

## Follow a spike

When a metric more than doubles for a while, Instatus highlights the window on the chart and says which service caused most of it.

If the metric is a request duration, **View N slow traces** opens [Traces](https://instatus.com/help/observability/traces) on the slow requests from that window. Open one to see which step was slow.

- [Dashboards](https://instatus.com/help/observability/dashboards): Put your key numbers on one page
- [Traces](https://instatus.com/help/observability/traces): See where each request spends its time
