Skip to content

Reference

Data sources

Every source the platform reads, what it contains, and what it's used for.

The platform reads six Azure sources and nothing else — no agents on your VMs, no network taps, no access to data inside your resources. The permissions reference maps each source to the role that grants it.

Each source is read through the two-role service principal: Cost Management Reader for the cost report, Reader for everything else.

SourceEndpoint · api-versionRoleWhat it’s used for
Azure Cost Details APIPOST …/Microsoft.CostManagement/generateCostDetailsReport · 2024-08-01 (ActualCost + AmortizedCost)Cost Management ReaderThe daily Cost API lane; dashboards, reports, audit baselines, savings verification
Azure Resource GraphPOST …/Microsoft.ResourceGraph/resources · 2022-10-01 (also AdvisorResources + ResourceChanges tables)ReaderResource inventory and metadata; mapping cost to what’s deployed; orphaned-disk detection
Azure AdvisorMicrosoft.Advisor/recommendations/read (via Resource Graph)ReaderAn input signal for findings, enriched with utilisation evidence
Azure Monitor metricsGET {resourceId}/providers/microsoft.insights/metrics · 2023-10-01ReaderIdle-VM detection and rightsizing confidence levels
Azure Activity LogGET …/Microsoft.Insights/eventtypes/management/values · 2015-04-01Reader”What changed” context behind cost movements
Azure Retail PricesGET https://prices.azure.com/api/retail/pricesNone (public)€ impact on findings; Reservation and Savings Plan coverage views

These sources power the whole free tier — dashboards, reports, chatbot, and the audit baseline — and populate after the first nightly run, within a day of connecting Azure.

For very large estates only, an optional export-based lane reads native FOCUS 1.2 parquet via azcopy from a storage container you control. It needs an extra storage grant and trails the daily lane on freshness, so it is documented separately in FOCUS exports and storage setup.

Ingestion runs once a night, fanning out across every subscription you’ve granted (10+ in a production estate). Each run moves through fixed stages:

TimeStageWhat happens
02:30AcquirePull from every source above
03:45DeriveShape raw data through the medallion warehouse (bronze → silver)
04:45BuildBuild gold tables and marts
05:45SyncRefresh the dashboards against the new data
06:15NotifyPost the pipeline summary card to your Teams channel

On first connection Infralign backfills up to 13 months of cost history, so dashboards and reports open with trend rather than a single day.

  • Data inside your resources: database contents, VM filesystems, blob data (other than the FOCUS export container), application logs, or secrets.
  • Anything requiring a write scope — none is ever granted.

Remove the role assignments or delete the app registration and the data flow stops immediately. See permissions for the revocation steps.

Your data is retained while the service is active. On offboarding, your tenant warehouse, the raw copies, and backups are deleted within 30 days, confirmed in writing. Hosting, per-tenant isolation, encryption, and sub-processors are covered in security and data handling.


Next: FAQ — the questions teams ask before connecting Azure.