Call analytics and real-time telemetry for per-user troubleshooting

What it does

Call analytics is the per-user, per-call view in the Teams admin center — every leg of a call or meeting with device, network and quality metrics, and flags on what went wrong. Real-time telemetry is the same idea for meetings still in progress, so the service desk can see a user's jitter and packet loss while they're on the line complaining.

Key facts

Both live in Teams admin center > Users > Manage users > select user > Meetings & calls.
Call analytics keeps the last 30 days of calls and meetings the user organised or joined. Users only show data if a media session actually happened.
Roles: Teams Administrator sees everything. Teams Communications Support Engineer (tier 2) sees full participant names plus the Advanced and Debug tabs (device names, IP addresses, subnet mapping, DNS suffix, SSID). Teams Communications Support Specialist (tier 1) sees only the searched user's name, other participants as "Internal User"/"External User", and no Advanced or Debug tab. Both support roles see phone numbers with the last four digits masked.
Yellow flags mean outside normal range but probably not the main cause; red flags mean a likely main cause. Sessions are marked poor for setup failures, audio network problems (loss, jitter, RTT, NMOS degradation, concealment) or device capture/render malfunction.
Real-time telemetry covers scheduled meetings and Meet now only — not PSTN calls, 1:1 calls or group calls.
With Teams Premium, real-time telemetry is collected automatically and kept for 7 days. Without it, telemetry is only collected while an admin has the in-progress meeting open in the admin center, and kept for 24 hours.
For meetings over 3 hours, only the last 3 hours of telemetry are available.
Supported on Windows, macOS, Android, iOS and the web client (including VDI 1.0 and 2.0), plus Teams displays, Teams phones, Teams Rooms and Teams Rooms on Surface Hub.
Metrics include jitter, packet loss, RTT, bitrate and codec for audio, video and sharing; frame rate, resolution and freeze counts for video and sharing; system and Teams CPU; battery; network type, Wi-Fi signal and band; and the user's mute/speaking state.
Not covered: view-only participants, anonymous users, attendees of live/large events (presenters only), CVI devices. External participants show as "unavailable". In federated meetings the organiser's tenant sees everyone; joining tenants see only their own users.

When to use / skip

Call analytics is the first stop for any individual complaint — "my calls were awful yesterday afternoon". Open the call, look for the red flag, and you'll usually find Wi-Fi, a Bluetooth headset, a VPN, or TCP fallback within a minute. Give the service desk tier 1 and tier 2 roles and train them on it; it removes a lot of noise from the network team's queue.

Real-time telemetry matters when you support high-value meetings live — board meetings, town hall presenters, executive calls — or run a service desk that takes "I'm in a meeting now and it's breaking up" calls. That's the strongest argument for Teams Premium on the support side: without it you only get data from the moment someone opens the meeting in the admin center, which is usually after the problem.

For trends and site-wide problems, don't use either. That's CQD's job; troubleshooting one user at a time won't show you that a whole building has the same issue.

Configuration decisions

Which support staff get Teams Communications Support Specialist versus Support Engineer. Engineer sees IP addresses, SSIDs and full names; that's a privacy decision as much as a technical one.
Whether to use administrative units to scope support staff to regions — and accept that AU-scoped admins lose EUII in CQD.
Whether the people you need automatic real-time telemetry for (executives, frequent presenters) justify Teams Premium licences. Check the tenant's own licence position rather than assuming.
Whether building data is uploaded, so call analytics can show the user's building and site rather than a bare subnet.
What the escalation path looks like: tier 1 triage in call analytics, tier 2 deep dive, then network team with CQD evidence.

Gotchas

No PSTN, 1:1 or group call support in real-time telemetry. People expect it for a Teams Phone complaint and find nothing until the call ends and call analytics catches up.
Without Teams Premium, real-time telemetry starts when you open the meeting. The bad five minutes before the ticket was raised are gone.
Session rating "Unavailable" means no QoE data was received — often a dropped call or a killed client. It isn't the same as "fine".
Users stuck in the lobby, view-only attendees and anonymous joiners never appear, so a guest's bad experience is invisible from your side.
Federated meetings: if the other organisation hosted, you'll only see your own users' legs.
Thirty days is short. If a VIP complains about a call from five weeks ago, there's nothing left.

Consultant notes

Write a one-page "how to read call analytics" crib sheet for the service desk: red flag, likely cause, first action. It pays back within a week.
Check the network type and Wi-Fi band first in nearly every case — 2.4 GHz Wi-Fi and TCP connectivity cover a surprising share of complaints.
For executive events, open real-time telemetry for presenters before the meeting starts if the tenant doesn't have Premium.
When you escalate to the network team, take the subnet, building and timestamps from call analytics, then the matching CQD view. They'll act on that.
If Teams Premium is already in the tenant for other features, make sure support teams know real-time telemetry is part of it.

Revisit if real-time telemetry ever extends to PSTN and 1:1 calls, or if Teams Premium's retention changes