FinOps for Fabric Capacity
Stop discovering Capacity Unit overruns at the end of the month. Real-time visibility, throttling early warnings, defensible chargeback, and a monthly executive view — all built on top of Microsoft's Capacity Metrics.
Microsoft's Capacity Metrics report is the raw data. This accelerator is the layer on top that makes it actionable. Extended semantic model with chargeback dimensions. Activator rules that page on impending throttling. A monthly exec report that ties capacity to business value. A playbook of cost-optimization moves ranked by impact.
Who this is for
- Fabric platform admins who've been throttled and want to never be surprised again.
- Finance / IT operations leads who need a credible chargeback model.
- Architects optimizing CU spend across multiple workloads.
- Anyone whose CFO has asked "what are we actually paying Microsoft for?"
What's in the box
| File | Type | What it does |
|---|---|---|
model/capacity_extended.bim | TMDL | Extension of Microsoft's Capacity Metrics semantic model with domain/owner dimensions, chargeback measures, and historical trending. |
config/chargeback_mapping.csv | CSV | Maps workspaces → domains → owners → cost centers. Edit once, drives all chargeback. |
activator/throttle_warning.json | JSON | Activator rule pack: warns at 70% sustained CU, alerts at 85%, pages at 95%. |
activator/runaway_workload.json | JSON | Activator rule pack: detects workspaces spiking 5x their baseline within 30 minutes. |
kql/historical_trending.kql | KQL | KQL queries against the Fabric audit log for long-window historical analysis. |
reports/executive_monthly.pbix | PBIX | Monthly exec report: spend by domain, top consumers, throttle events, ROI commentary placeholders. |
reports/admin_realtime.pbix | PBIX | Real-time admin dashboard: current CU%, top workspaces, incoming pipeline schedule. |
tools/fsku_rightsize.xlsx | XLSX | F-SKU right-sizing analyzer: input usage profile, output recommended SKU + projected savings. |
playbook/cost_optimization.md | MD | 15 ranked optimization moves with estimated CU savings. The "what to do this quarter" list. |
playbook/throttle_response.md | MD | The 4-step playbook when an alert fires: triage, isolate, mitigate, post-mortem. |
Top 5 optimization moves from the playbook
- Move Import-mode semantic models to Direct Lake where eligible. Saves 30–60% of model-related CU.
- Throttle Power BI report-page auto-refresh. Most "real-time" reports don't need 5-second refresh; 60s is fine.
- Schedule heavy Spark jobs off-peak. Same total CU but different shape — moves you under the bursting ceiling.
- Consolidate duplicate semantic models. Most large estates have 3–5 versions of the "same" model. Pick one, kill the rest.
- Right-size your F-SKU using the analyzer. Most teams over-provision by one SKU "just to be safe."
The full playbook has 15 moves with estimated savings and effort. Most clients see 20–40% capacity-spend reduction within the first quarter.
Pricing
Frequently asked questions
Will this conflict with Microsoft's Capacity Metrics report?
How accurate is the chargeback?
Can I customize the chargeback model?
What about Premium per User (PPU) licenses?
Refund policy?
Never get throttled by surprise again
Most clients recoup the cost of this accelerator in the first month of CU savings.