Dashboards

Three Layout Families - Dashboards, Timeboards, and Screenboards

When you create a new dashboard, the layout you choose comes in three families (as of August 2026; source: Dashboards). The standard Dashboards layout is grid-based, and lets you place a wide variety of objects such as images, graphs, and logs within a frame up to 12 grid columns wide. Timeboards use an automatic layout in which the entire screen is locked to a single time axis, which makes them suited to troubleshooting where you need to line up multiple indicators at the same moment. Screenboards use free-form placement and are used for real-time status boards and presentation screens for reporting. The refresh interval of the screen is tied to the time range being displayed; for example, a 1-minute view refreshes every 10 seconds and a 1-day view every 3 minutes. In practice, starting with the standard grid layout is enough, but for the specific case of an investigation screen where it matters that the time axis is aligned across all widgets, keeping a Timeboard-style approach in mind makes it harder to confuse cause and effect in time.

Choosing Widgets - Time Series Is Not the Only Answer

The flagship widget is the time series graph (timeseries), but the basic set that the official learning course introduces first also includes the query value, which shows a single number in large type; the top list, which lines up the top N items; the table; the distribution, which shows how values are spread; and the pie chart (as of August 2026; source: Dashboards). The axis for choosing between them is what the reader needs to decide. If you want to know at a glance how far you are from a threshold, a query value is faster than a time series; if you want to know which host stands out, a top list is the shortest path. There are also SLO widgets that show the remaining error budget of an SLO and architecture-type widgets for drawing configuration diagrams, and if your team operates SLOs, keeping the remaining budget permanently visible in a dedicated widget brings the question of whether we can take a risk right now into everyday conversation. Before adding another widget, asking yourself whether you can write in one line the question this screen is meant to answer is, in the end, what works best.

Template Variables - Making One Dashboard Do the Work of Dozens

If you duplicate a dashboard with the same layout for every service and every environment, you end up fixing all of them every time you make a change. Template variables are the official answer to this duplication nightmare: define a tag key (for example, env or service) as a variable, and switching the value in the drop-down at the top of the screen switches the filter on every widget at once (as of August 2026; source: Template Variables). The default value is * (everything), and frequently used combinations can be stored as saved views. The candidate values that appear in the drop-down are populated from the data sources used by the widgets on that dashboard and the displayed time range, so when a value does not appear in the list, a common misunderstanding is that the variable is broken; usually there is simply no data with that tag value in that period. The precondition for template variables to show their true value is a consistent tagging convention, and this dependency is covered in detail in our guide to tagging.

Public URLs, Invite-Only Sharing, and the Audit That Completes the Job

Dashboards can be shown to people who are not logged in. There are four families of sharing: shared dashboards that issue a public link, embed codes for individual graphs, sharing a widget as an image, and scheduled email reports (as of August 2026; source: Sharing). Public links come in two kinds, viewable by anyone or invite-only; with invite-only links, you grant access per email address or per email domain, and you can set an expiration date for each invitation. When the date arrives, access expires at midnight local time. What to watch out for is that public links tend to get pasted around as if they were internal URLs. Metric names and host names are themselves internal information, and a publicly shared screen keeps refreshing every 30 seconds regardless of its time range. The Public Sharing page in your organization settings lists the public dashboards and public graphs in the organization, so the sharing feature only runs safely once you pair it with a quarterly audit in which you open that page and revoke public links that nobody looks at anymore.

JSON Management - Duplicating, Managing as Code, and Recovering from Accidental Deletion

A dashboard's definition can be copied, imported, and exported wholesale as JSON (as of August 2026; source: Configure Dashboards). What this means is that you can turn a dashboard from a hand-made artifact into a configuration object managed as code. Put the JSON in git and you have a change history and the ability to review changes. When you stand up a new environment, importing the JSON brings up the same screen in minutes. Note, however, that import is an operation that overwrites all existing content, so beware of the accident of pouring old JSON into a dashboard that is being edited jointly. If you delete a dashboard by mistake, you can restore it from the Recently Deleted list, and it is permanently deleted 30 days after it lands there. Just knowing that the cry of an important dashboard has vanished comes with a 30-day grace period changes your first response.

How to Design a Screen That Actually Gets Looked At

There is nothing technically hard about arranging widgets. What is hard is having the screen you built still be opened three months later. The typical failure is the dashboard that contains everything that could be put in it; a screen optimized for nobody's decision about anything ends up opened by nobody. As a prescription, we propose splitting into three dashboards by purpose.

Each board survives because you can say who opens it and when. If you embed a direct link to the investigation board in the notification message of a monitor, the investigation starts one click away from the alert, and the dashboard changes from a place you go to look into a place you open when called. Once this takes hold, stale boards reveal themselves naturally: a board that is not opened is a board that has dropped out of the flow.

Dashboards versus Notebooks

A similar tool is the Notebook. A Notebook is a rich-text document that multiple people can edit together, and you can embed live Datadog graphs directly in the middle of the text. Datadog lists incident investigation records, postmortems, and runbooks as its main uses, and provides templates such as Incident Report and SLO Specification (as of August 2026; source: Notebooks). The dividing line can be drawn like this: a dashboard is a standing screen for looking at the same question every day, while a Notebook is a document for preserving a single investigation or explanation together with prose. Rather than continually pasting screenshots of an incident investigation into Slack, pasting them into a Notebook in chronological order gives you postmortem material as-is. Conversely, if you start using a Notebook as a substitute for standing monitoring, that is the signal to promote it to a dashboard. When you can operate by moving back and forth between the two, investigative insight stops being disposable and accumulates in the organization.