---
title: "Cluemetry"
description: "Cluemetry, Curade Mini's reasoning engine, explains why a reading, a problem or an item looks the way it does, from what monitoring recorded about that moment."
url: https://curade.ai/en/curade-mini/docs/features/cluemetry/
product: "Curade Mini"
version: "1.0.0"
breadcrumbs:
  - "Curade Mini docs"
  - "Features"
  - "Cluemetry"
---

# Cluemetry

Cluemetry, Curade Mini's reasoning engine, explains why a reading, a problem or an item looks the way it does, from what monitoring recorded about that moment.

**Cluemetry** is the app's reasoning engine. It explains why something reads the way it does, from what monitoring recorded about that moment, and shows the facts each finding stands on. It is not AI: it follows set rules over monitoring's own facts, so the same question about the same moment gets the same answer. It carries a **BETA** badge while it is still being tested.

![Response time reading picked on a chart, with Cluemetry's Why sheet tracing the climb to a busy host and naming a deployment](https://curade.ai/_astro/07-chart-why.xk1xYcxl_2uuMK8.webp "Tap any spike to see why.")

## Where it shows up

- **Why?** under a reading you pick on a chart, and on a chart's caption when some intervals have no reading.
- **What's behind these numbers** on a service, host or application, and **Why this happened** on a problem.
- **What drives this number** in a dashboard widget's sheet.
- The short reason on a list card that needs attention, and the line above a list when several items share a cause.
- CLEO's answers when you ask why something reads the way it does, shown as **Cluemetry Analysis** while it works.

## What it looks at

Cluemetry works on the exact moment on screen: the interval you picked, the page's time window or the problem's own time. Around it, it gathers:

- the problems open at the time, and what monitoring named as their root cause;
- deployments, restarts and configuration changes from shortly before;
- other events monitoring recorded on the item, its hosts and its processes;
- whether the hosts and processes behind the item were up and reporting;
- what the item depends on, and what depends on it;
- how the same reading looked at the same time yesterday and last week, or just before;
- your own monitoring alerts on the item.

It reads only what your role can see on its own pages. It never reads logs, and it never changes anything in monitoring.

## How it labels a finding

Something that happened at the same time is not always the cause, so every finding carries a label saying how closely it is tied to the item:

| Label | What it means |
|---|---|
| **Root cause, named by monitoring** | Monitoring itself named this as the root cause of a problem at the time. It is the only kind of finding Cluemetry calls a cause. |
| **Recorded on this item** | Something monitoring recorded on the item itself, such as a deployment or a restart on it. |
| **Related to this item** | Something on what the item depends on, at the same time, such as its host running short of CPU. |
| **Possibly unrelated** | Something that happened at the same time with no known link to the item. |
| **Worth knowing** | Background, such as how the same reading looked yesterday. |

An explanation opens with a one line verdict, then the findings strongest first, the changes and problems around the moment, what was checked and found normal, and what could not be checked and why. Something it could not check is unknown, never a sign that all was fine.

## Intervals with no reading

A gap in a chart is drawn as a gap, never as a zero. Cluemetry says why it is there, leading with the first reason that holds: the measure is not collected in the environment, the interval is still arriving, the item did not exist yet or had stopped reporting, the hosts behind it were down, monitoring itself was unavailable, or the service simply received no requests. A reading that covers only part of its interval is marked as partial.

## Good to know

- An explanation answers about the exact moment on screen, so it stays true for a time window in the past.
- An explanation of something still unfolding, such as an open problem, can change as monitoring catches up, so asking again later may add to it.
- **Ask CLEO about this** hands an explanation to CLEO, which can talk it through or look further ([CLEO](https://curade.ai/en/curade-mini/docs/features/cleo/)).
- Nothing is explained while the environment is suspended or its access token has lapsed.