
This article is for Azure administrators, FinOps teams, DevOps teams, and cloud leaders. It covers what Advisor evaluates, what it can't see, how to validate its findings before acting, and where block storage falls through the cracks.
Key Takeaways
- Advisor analyzes configuration and usage across five categories: Cost, Security, Reliability, Operational Excellence, and Performance.
- It's a starting point, not a replacement for Azure Monitor, Cost Management, or application observability.
- Validate low utilization and projected savings against peak demand, criticality, and ownership.
- Block storage can stay over-provisioned even when your Advisor dashboard looks healthy.
What Azure Advisor Catches
Azure Advisor analyzes your resource configuration and usage telemetry, then applies Microsoft best-practice logic to produce personalized recommendations. Coverage includes VMs and scale sets, AKS, App Service, SQL and Cosmos DB, storage accounts, and networking.
Recommendations refresh daily, and an implemented change can take a day to register.
Cost recommendations
Cost recommendation families include:
- Resizing or shutting down underutilized VMs and scale sets
- Unused resources, such as unattached disks (preview) and gateways with no connections (preview)
- Reservation and savings plan purchases, including managed disk reservations
- Storage tier changes, such as Premium storage for high transaction-per-TB ratios
The VM logic is specific. A shutdown recommendation needs P95 maximum CPU below 3% and outbound network utilization below 2% over seven days. Memory isn't considered.
Resize recommendations check the proposed SKU. For user-facing workloads, that means P95 CPU and network at or below 40%, and P99 memory at or below 60%.

Performance, reliability, security, and operations
Advisor also flags patterns tied to bottlenecks, availability risks, security weaknesses, and operational gaps. Common examples include:
- OS disks approaching limits, with HDD-to-SSD upgrade suggestions
- Premium SSD v2 data disks instead of P-series for new deployments
Its Service Retirement workbook gives resource-level retirement visibility for migration planning.
Prioritization and action guidance
The portal offers category tiles and filters by subscription, resource group, and resource type. CSV and PDF exports respect those filters. You can postpone a recommendation (it returns later) or dismiss it until reactivated. Some recommendations carry Quick Fix actions, but availability varies by recommendation.
What a recommendation actually tells you
Keep four things separate:
- A detected condition
- A suggested change
- An estimated benefit
- A completed remediation
Only the first is observed. VM savings estimates use retail rates, exclude negotiated discounts, and ignore existing reservations or savings plans. Actual savings can be lower, and some cross-series changes can raise cost.
What Azure Advisor Misses
Application behavior and root cause
Advisor works from configuration and usage signals. Diagnosing why an application is slow takes distributed tracing, dependency maps, and log analysis, which is what Azure Monitor and Application Insights provide. Treat Advisor as a prompt to investigate, not a diagnosis.
Business context and ownership
Advisor can't know a quiet VM is a disaster recovery standby, a licensing anchor, or a batch server that wakes up quarterly. Microsoft's own guidance lists upcoming traffic, unmeasured requirements, and DR scenarios as considerations before acting. Tags close the gap:
- Accounting tags (department, cost center)
- Ownership tags (responsible team)
- Approval workflows before you act on a recommendation
Short views of demand
The default VM lookback is 7 days, configurable to 7, 14, 21, 30, 60, or 90 days. Changes can take 48 hours to affect recommendations. A week of data misses month-end processing, quarterly reporting, and seasonal peaks. Extend the window to 30 or 90 days for anything with a calendar rhythm.

Granular storage utilization
Advisor documents unattached disk review (preview) and some disk-type changes. Its cost and performance catalogs don't document:
- Capacity rightsizing for attached disks
- Zero-I/O detection
- Mounted or unmounted state inside the guest OS
Knowing a disk exists differs from knowing whether it's mounted, active, reserved on purpose, or safe to resize.
Architecture, multicloud, and organization-wide gaps
Advisor isn't an architecture review. It won't redesign workloads, consolidate services, inventory AWS and Google Cloud, or map waste to business units. Cost Management allocation helps with the last one, though redistributing costs doesn't change invoices or billing responsibility.
Azure Advisor vs. Azure Monitor and Azure Cost Management
Monitor versus Advisor
Azure Monitor collects and analyzes metrics, logs, and traces for observability, alerts, and diagnostics. Advisor interprets selected configuration and usage signals into recommendations. Mature teams use both: Monitor explains what's happening, Advisor suggests what to change.
Cost Management versus Advisor
Cost Management handles spend analysis, budgets, forecasting, anomaly alerts, and allocation. Advisor cost recommendations even appear inside it. A practical workflow: Cost Management sets the financial baseline, Advisor identifies candidate changes, and Cost Management confirms the bill moved.
Use the right tool for the question
| Question | Tool |
|---|---|
| What best-practice or resource change should I consider? | Advisor |
| What is happening, and why? | Azure Monitor / Application Insights |
| Where is spend going, and did the change affect the bill? | Cost Management |
Security and architecture questions often need dedicated security tooling or expert review on top.
How to Use Azure Advisor Safely and Effectively
- Baseline first. Record current spend, utilization, performance, availability objectives, commitments, owners, and criticality using Cost Management and Monitor.
- Triage by risk. Start with clearly orphaned resources, then non-production rightsizing, then production. Weigh estimated savings against outage, compliance, and rollback risk.
- Validate with longer evidence. Compare recommendations against peak load, scheduled jobs, seasonality, and SLOs, using a lookback longer than the default where needed.
- Route through ownership. Confirm the owner, record the reason for any exception, and require approval. Well-Architected guidance calls for access controls, audit trails, and infrastructure defined in version-controlled code. Keep a rollback plan for production changes.
- Measure realized results. Compare post-change cost and performance to the baseline. Separate realized savings from estimates.

Schedule recurring reviews. Postponed recommendations return to Active, but dismissed ones stay hidden until someone reactivates them, so they can quietly vanish from governance.
The FinOps Foundation framework frames this as Inform, Optimize, Operate, a loop rather than a one-time cleanup.
The Cloud Storage Blind Spot and Where a Complementary Platform Fits
Why block storage stays inefficient
Teams over-provision disks to protect performance and uptime. Then they rarely revisit them. Managed disk capacity charges follow provisioned size, not bytes written, and orphaned disks keep billing. Azure doesn't support in-place disk shrinking either. Growing is easy; shrinking means migrating data and risks loss.
Capacity, performance, attachment, mount status, and I/O activity are separate questions. A dashboard that answers one doesn't answer the rest.
Gartner's storage cost research names overprovisioning, poor lifecycle management, and manual controls as cost drivers. Lucidity's own figures put average enterprise disk utilization near 30%, with 8-10% of disks tied to downtime.

Where Lucidity can complement Advisor
Lucidity is a block storage optimization platform for enterprises that want real-time visibility and autonomous management across Azure, AWS, and Google Cloud. It works as an application-agnostic, no-code layer. Its AutoScaler expands and shrinks block storage based on real workload behavior, with no infrastructure changes.
Its Lumen product identifies four idle-disk types: unattached, reserved, unmounted, and zero-I/O. These can represent up to 70% of unused block storage spend and don't appear in native cloud dashboards. Cleanup is one-click, auditable, and reversible. A free Assessment shows disk utilization and potential savings before any commitment.

How to evaluate an additional layer
Ask any vendor:
- Does it cover every relevant volume?
- Can it tell idle disk states apart?
- Does it account for performance and workload risk?
- Does it work across subscriptions and clouds?
- What approvals, audit controls, rollback, and realized-savings reporting exist?
Microsoft's FinOps guidance points the same way. Map tools to organizational needs, confirm they complement existing coverage, test before scaling, and reassess periodically.
Frequently Asked Questions
What are the key differences between Azure Monitor and Azure Advisor?
Azure Monitor collects, visualizes, and alerts on telemetry. Azure Advisor analyzes selected resource signals and returns personalized optimization recommendations.
What is the purpose of Azure Advisor?
Advisor recommends actions to improve cost efficiency, performance, security, reliability, and operational practices across your Azure resources.
What are Azure Advisor's recommendation categories?
The categories are Cost, Security, Reliability, Operational Excellence, and Performance. Anything else, such as budgeting or log analytics, belongs to another service like Cost Management or Monitor.
Is Azure Advisor free?
Microsoft's pricing page says you can access recommendations at no additional cost. Related services, automation, or third-party platforms may carry separate charges.
Can Azure Advisor automate remediation?
Not fully. Some recommendations offer portal-only Quick Fix bulk actions, while CLI, PowerShell, REST, and Resource Graph support scripted workflows. Remediation still depends on permissions, approvals, testing, and rollback.