Before you begin
- A Professional plan or above: all four pages are tier-gated and the server enforces it. A few tabs need Enterprise, as noted below
- The matching view permission. End-User Experience and Environment Insights need ai.environment_insights.view, Session Intelligence needs intelligence.session_engine.view, and Optimization needs ai.optimization_engine.view plus the pool-management permission (Administrator or Operator)
End-User Experience Score
Open Atlas Operations → Compass → End-User Experience. The page scores the experience your users actually had on a 0–100 scale, where lower means worse, and lists the worst scores first so the people and pools that need attention stand out. Pick a time window of the last 7, 14 or 30 days, and use Export CSV to take the list with you.
Each score is built from these factors, measured over the window:
- Logon duration (95th percentile): 15 seconds or less scores full marks, 60 seconds or more scores zero
- Round-trip latency (95th percentile): 50 ms or less is good, 250 ms or more is bad
- Input lag (95th percentile) and frame rate (average), when the session reports them
- Disconnect rate, reconnect rate and region-failover events
Ratings follow the score: Good is 80 and above, Fair 60–79, Poor 40–59 and Critical below 40. The At risk count on the page is everything below 60.
How the score is calculated
Each factor is scored from 0 to 100 between its good target and its bad threshold, then the composite is the weighted average of the factors that had data. A factor with no telemetry in the window is left out and its weight is shared across the others, so the applied weights always add up to 100%. Click any environment to open its breakdown: every factor's measured value, its good-to-bad target, the weight applied, and how many points it contributed to the score.
What N/A means
N/A means Atlas did not have enough experience telemetry in the window to score that environment or user. It is not the same as a score of 0, which would be a real, measured bad experience. A score needs at least one of logon duration, latency, input lag or frame rate; disconnect and reconnect rates on their own are never scored, because a handful of sessions with a burst of disconnects says little about how the session felt. Widen the time window, or check back once sessions have run and their telemetry has arrived.
By environment and By user
The By environment tab scores each host pool, with a score distribution you can click to filter to one rating band. The By user tab scores each person in the selected tenant, with their last environment and session count, and an Access column that shows how they reached their desktop:
- Portal: their sessions were launched through the User Portal
- Mixed: some sessions came through the User Portal and some did not
- Not via portal (direct): they ran sessions but none started from the User Portal, so nothing the portal enforces or measures applies to them. This is flagged as a finding to review, not a fault
- Unknown: User Portal launch history could not be read for the period, so Atlas cannot say either way
Filter the user list by access path to find everyone connecting outside the User Portal, then hover the Access badge to see how many portal launches succeeded against the sessions they ran.
Environment Insights (Environment Engine)
Open Atlas Operations → Compass → Environment Insights. The page, titled Environment Engine, brings health monitoring and root-cause analysis for the selected tenant together in six tabs:
- Overview: an environment health score with the meters behind it (VM health, session health, FSLogix health, infrastructure, Azure quota and app credentials), headline counts and anything pulling the score down
- Alerts: real-time operational alerts such as high CPU or memory, slow logins, slow FSLogix, disk latency and low disk space, which you can search, filter by severity, type and time range, and export
- Incidents and Findings: root-cause analysis incidents and findings with their evidence and recommendations. These two tabs need the Enterprise plan
- Map: an explorer of your tenant, subscriptions, resource groups, host pools, session hosts and networking, with health colors, dependencies and blast radius. It needs the pool-management or audit-view permission
- FSLogix: FSLogix performance insights for your profile storage
Session Intelligence (Session Engine)
Open Atlas Operations → Compass → Session Intelligence. The page, titled Session Engine, covers live user sessions and session recovery in four tabs:
- Overview: active, disconnected and stuck sessions, recoveries today, the busiest host pools and recent recovery events
- Active Sessions: every live session with its user, host pool, session host, state and start time
- Recovery Events: the history of automatic and self-initiated session recoveries, with completed, failed and in-progress counts and the success rate
- Self Recovery Stats: recoveries users started themselves from the self-service portal, with the success rate and unique users
Without the permission to view full user details (held by Administrators and Operators), user names, display names and IP addresses are masked and a banner says so. Session replays have moved to their own page, Microsoft 365 → Entra → Session Replay; the Overview still shows how many replays are available and links there.
The Optimization hub
Open Atlas Operations → Compass → Optimization. The page, titled Optimization Engine, gathers the cost and efficiency engines in one place:
- Overview: idle hosts detected, active disk downgrades, density mode and estimated monthly savings
- Smart Disk Tiering: moves a deallocated session host's OS disk from Premium SSD to Standard HDD and back to Premium SSD when the VM starts. Only host pools with Start VM on Connect are eligible, and you can run it automatically, on a schedule only, or manually
- DrainGuard: deletes idle session hosts during off-hours with maintenance-mode safeguards. See DrainGuard Setup
- Adaptive Density: raises and lowers each pool's max sessions per VM with demand. This tab appears only on the Enterprise plan. See Configuring Adaptive Density