An annotation is a dated note that renders on your charts: a release shipped, a sale started and ended, tracking broke for a day. Every chart with a time axis shows the annotations that fall in its date range, so when a metric moves, the candidate explanations are already drawn on the chart instead of living in someone’s memory of the release calendar.

This is the habit that pays off later: causal decomposition tells you which lever moved a KPI, and annotations tell you what you did with that lever that week. Teams that annotate consistently answer “what happened here?” in seconds.

What an annotation carries

A date, or a range with a start and end (a sale, an event, an incident window).
A short title and optional description.
Tags you define, so annotations can be grouped and filtered by kind: releases, live-ops, marketing, whatever fits your studio’s vocabulary.

How they render

Annotations appear as markers on the chart’s time axis; ranges render as shaded bands. Hover shows the details. Placement is breakdown-aware: on charts with comparison periods or grouped bars, the marker attaches to the right occurrence of its date instead of smearing across the chart.

Who can annotate

Creating and editing annotations requires the Editor role or above; every member sees them. Annotations are shared, curated context, so they sit on the team side of the personal-vs-shared line.

Annotations from AI chat or CI

The AI chat integration exposes create-annotation and list-annotations tools, with the same Editor+ requirement. Two useful consequences:

In chat, “annotate yesterday’s dip: tracking outage on Android” works.
A release pipeline can post an annotation automatically on every deploy, using a personal access token against the same tools, so release markers appear on charts without anyone remembering to add them.