Microsoft Teams Monitoring: Call Quality | Cloudmon

Microsoft Teams Monitoring

Cloud

Microsoft Teams Monitoring

Track Teams call quality across your Microsoft 365 tenants, by site, user, and call, and see whether a poor call comes from your network or from a Microsoft service issue.

Overview

Microsoft Teams monitoring reads call records from Microsoft Graph and shows how Teams calls performed in each tenant. Cloudmon polls each tenant every 15 minutes from a Linux probe, using an app registered in Microsoft Entra. Microsoft publishes a call record 15 to 30 minutes after the call ends, so results are near real-time. This article covers monitoring of Teams calls. To send Cloudmon alerts to a Teams channel, see Microsoft Teams integration.

Microsoft Teams monitoring: how to set it up
STEP 1
Register app
Register an app in Microsoft Entra and grant it the Graph permissions
STEP 2
Add tenant
Add the tenant in Cloudmon with the app details and a probe
STEP 3
Map sites
Map your subnets to named locations in the Site Map
STEP 4
Review calls
Follow quality by site, user, and call

Prerequisites

RequirementDescription
ProbeA Linux probe with internet access to graph.microsoft.com.
Entra app registrationAn app registered in Microsoft Entra. Keep its Directory (tenant) ID, Application (client) ID, and a client secret Value. For the steps, see Microsoft's guides to registering an app and adding a client secret.
Graph application permissionsCallRecords.Read.All to read call records, and ServiceHealth.Read.All to list issues in the Microsoft Issues tab. For the steps, see Microsoft's guide to adding application permissions to Microsoft Graph, which also covers admin consent.

Adding a Tenant

  1. Navigate to Cloud → SaaS → Microsoft Teams and click Add.
  2. Fill in the fields as described in the following table.
  3. Click Test Connection to check the credentials and Graph permissions from the probe.
  4. Click Save.
FieldDescription
NameRequired. The display name for this tenant in Cloudmon.
ProbeRequired. The Linux probe that polls this tenant.
Directory (tenant) IDRequired. Find it in Entra → App registrations → your app → Overview.
Application (client) IDRequired. Find it in Entra → App registrations → your app → Overview.
Client secretRequired. The secret Value from Certificates & secrets, not the Secret ID. It is stored encrypted. When editing a tenant, leave it blank to keep the current secret.
Secret expiryThe date the secret expires. Cloudmon uses it to raise an alarm before the secret expires.
Alarm RuleThe thresholds and severities for this tenant. Default Alarms for Microsoft Teams is selected by default.
User GroupsThe user groups to notify when alarms are raised.
Third Party ServicesThe third-party integrations for alarm notifications.

Tenants Page

The Microsoft Teams page summarizes all tenants by state, and charts calls, poor call percentage, and average MOS (Mean Opinion Score) by tenant over the last 24 hours. The Microsoft Teams Tenants table lists each tenant.

ColumnDescription
NameThe tenant. Click it to open the tenant page.
StateThe monitor state, such as Critical.
Tenant IDThe directory (tenant) ID.
ProbeThe probe that polls the tenant.
StatusWhether monitoring is active.
Calls (24h)The calls in the last 24 hours, with the number that failed.
Poor % (24h)The share of calls that were poor.
MOS (24h)The average estimated Mean Opinion Score.
Last PolledWhen the probe last polled the tenant.
Secret ExpiryThe client secret expiry date.
ActionsTurn monitoring on or off, Poll Now, Subnet → Site Map, edit, and delete.

How Call Quality Is Measured

A call is poor when at least one of its streams crosses one of the following thresholds.

MetricPoor when
JitterOver 30 ms.
Packet lossOver 10%.
Round-trip time (RTT)Over 500 ms.
Concealed audioOver 7%.
Microsoft Graph does not report MOS directly. The MOS shown in Cloudmon is an estimate, calculated from RTT, jitter, and loss on a scale of 1 to 5.

A MOS of 4.0 and above is good, and a value under 3.5 is poor. The average for a tenant uses the worst audio leg of each call.

Tenant Details

Click a tenant to open its page. Use the time range control to change the period, and the page menu for Poll Now, Subnet → Site Map, and Edit.

Overview

ItemDescription
CallsTeams calls that started in the selected range, split into peer-to-peer and group. Click to open the calls.
Healthy CallsThe share of classified calls where no audio, video, or screen-share stream crossed a poor threshold.
Poor CallsThe share of classified calls with at least one poor stream.
Failed CallsThe share of calls that failed during setup or dropped mid-call.
Avg MOS (est.)The average estimated MOS.
Avg Jitter / Loss / RTTThe average network figures for the range.
Active UsersDistinct signed-in users, org users and guests, who joined at least one call. Phone and anonymous participants are not counted.
Call HealthGood, poor, and failed calls over time.
Average MOS (estimated)The estimated MOS over time.
Quality by ModalityThe share of audio, video, and screen-sharing streams with no poor metric. Screen sharing is judged on loss only, since low frame rates are normal for static content. Click a row with poor streams to open those calls.
Top Poor-Call CausesThe most common reason each poor call was flagged, such as packet loss, jitter, weak Wi-Fi, media over TCP, or device glitches. Click a cause to open those calls.
PlatformsParticipant legs by Teams client platform, such as Windows, macOS, web, iOS, Android, and Teams Rooms.
Connection TypesParticipant legs by network: wired, Wi-Fi, mobile, or tunnel/VPN.

Sites

Summary cards show the number of sites, with those that have too many poor calls, and the average latency, jitter, packet loss, and UDP share. The Site Statistics table lists each location with its subnets, users, calls, good and poor calls, poor percentage, MOS, and inbound and outbound latency, jitter, and loss, along with its UDP share.

  • Unmapped means the subnet is not in the Site Map. Add it to name the location, as described in Mapping Subnets to Sites.
  • Unknown network means Microsoft did not report network details for those participants, which is common for phone, bot, and some mobile legs.
  • UDP % is the share of participants whose media used UDP. Teams falls back to TCP when UDP ports 3478 to 3481 are blocked, which degrades quality. When over 20% of media at a site falls back to TCP, open UDP 3478 to 3481 to Microsoft 365 from that site.

Users

The table lists everyone who joined a call in the selected range, with the most poor calls first. Search by name or email, or filter by All, Org Users for users of your Microsoft 365 tenant, or Guests & External for guests, federated users from other tenants, bots and apps, and phone participants.

Each row shows the site, IP address, connection, protocol, calls, poor calls, MOS, platform, audio device, and when the user was last seen. Click a name to open their platforms, connections, Teams version, sites, IP addresses, subnets, audio devices, and recent calls.

Calls

The table lists each call with its health, type, participants, number of participants (No. of Participants), start time, duration, MOS, poor cause, and sites. Search by participant, organizer, or call ID, or filter by health and type. Health is Good, Poor, Failed, or Unclassified, and type is P2P or Group.

Click a call to open its details, as described in the following table.

SectionDescription
SummaryThe health, type, date and time, number of participants, organizer, lowest MOS, and call ID.
ParticipantsOne row per participant leg: client version, location, IP address and subnet, reflexive IP, relay IP, connection, protocol, audio received and sent as jitter, loss, and RTT, MOS, impact, and device. Click Calls beside a participant to open their calls.
TimelineQuality over the call for each participant, marked good, poor, or no metrics. Switch between Received and Sent, and between Audio, Video, and Screen Sharing.

Microsoft Issues

This tab lists Microsoft Teams service issues from the last 30 days, with their status, ID, title, type, feature, start time, and last update. Click an issue to read its details and user impact. When an issue is active, a banner warns that poor calls in the period may not be caused by your network. The tab needs the ServiceHealth.Read.All permission.

Mapping Subnets to Sites

Until subnets are mapped, calls are grouped by subnet. Map them so that the Sites tab shows named locations.

  1. On the tenant page, open the page menu and click Subnet → Site Map.
  2. Click Add Row and fill in the fields as described in the following table, or click Import CSV to add many rows at once.
  3. Click Save.
FieldDescription
LocationRequired. The name of the site.
SubnetRequired. A CIDR such as 10.20.0.0/16 covers a whole range. Without a mask, a value such as 192.168.1.0 matches that subnet exactly as Teams reports it in the Sites tab.
CountryThe country code of the site, such as IN.
LongitudeA number from -180 to 180.
LatitudeA number from -90 to 90.

A CSV file has the columns Location, Subnet, Country, Longitude, and Latitude. The header row is optional. Click Export to download the current map.

Alarms

To configure alarm rules, see How to Configure Alarm Rules.

Reports

To create a report, see Reports.