Creating tiles

A tile answers one question with one query against one data source. Each recipe in this section builds a single tile end to end, and they are grouped by the kind of metric they produce.

Recipes by metric type

Each group collects the recipes for one family of metric, with the shared definitions and calculations up front.

  • Alert metrics: Five recipes covering alert volume, out-of-hours distribution and actionable alert rate.
  • Deployment metrics: Eight recipes covering deployment frequency, lead time, change failure rate and deployment risk.
  • Incident metrics: Ten recipes covering MTTA, MTTR and how far an incident spread.

What a recipe covers

Step
What it decides
Data source and data stream
Which connection to select, and which stream within it returns the right shape of data.
Parameters
The query itself, plus any filters deciding which records come back. This is the part that changes most between tools.
Columns
How the returned data is grouped, aggregated and typed. A metric usually exists here rather than in the query, as an aggregation or a calculated column.
Visualization
The chart type, and which column maps to which slot.
Monitors and KPIs
Where a tile carries a threshold worth watching or a value worth reusing elsewhere.

What every recipe assumes

That the data source it names is already connected, and that you can open the tile editor and run a query. See exploring data for the tile editor itself.

Recipes do not repeat the connection steps. If the data source a recipe names is missing from the list in the tile editor, it has not been added to your organization yet.

Adapt a recipe to your own tools

Each recipe is written against one specific data source, but the metric behind it is rarely tied to that tool. A count of successful deployments is the same measurement whether the records come from Azure DevOps, GitHub Actions or Jenkins.

To run a recipe elsewhere, keep the shaping and the visualization and substitute the data stream and the field names. Where a recipe depends on a field only one tool provides, such as a service desk SLA metric, it says so and gives an expression that derives the same value from raw timestamps instead.

Was this article helpful?


Have more questions or facing an issue?