Estimated monthly savings if every recommendation in the category is applied.
Savings by disruption level
How much is on the table before anything running is touched.
Actual spend by service
Billed spend from Cost Management.
What you have already bought
Paid for and not being used
What the AI cost
The meters behind it
What Copilot cost
The meters behind the metered half
Contain and modernise
Guardrails: what stops the waste coming back
Moves: newer services and SKUs that change the shape of the bill
Deadlines: retirements and end of support for what is running
Application signals
How to read this. Each row of cells is one weekday, each cell one hour, averaged over the window: the more solid the cell, the busier the hour. The line above each grid counts the hours under 30% of the plan's own peak. A plan that is pale for most of the week is paying for instances it does not need all week; where a rule could price that, the finding is on the Actions tab (load-shaped autoscale or a smaller tier). A grid with no shading means the plan received no requests at all.
Weekly load per App Service plan
Cost per thousand requests
Who calls what
Where the money goes
The most expensive resources last month, from Cost Management.
Spend trend
Spend by resource group
Who owns the bill.
Since last scan
Partial data. Some sources could not be collected, so savings may be understated:
Before you run anything. Nothing in this report has been run. Every command is here for you to review, test outside production and run yourself, in a change window you choose. Take a backup or snapshot where the way back needs one. Savings are estimates until a later bill confirms them.
What isn't being used
Every metered resource with a plain verdict from Azure Monitor, including the ones no rule fired on. Idle and empty things first; the cost column is last month's actual bill for that resource.