Troubleshoot audio quality issues by using QualityMetrics in the contact record - Amazon Connect

Troubleshoot audio quality issues by using QualityMetrics in the contact record

Important

The topics and content in this section are for IT administrators who have experience with investigating network and telephony issues.

You also need to be familiar with how to access data in an Amazon Connect contact record.

Where to find QualityMetrics

Amazon Connect provides QualityMetrics in the contact record for each connected call.

QualityMetrics is part of the contact object that you get as a response when you call the DescribeContact API. The following snippet from the DescribeContact response shows what it looks like:

"QualityMetrics": { "Agent": { "Audio": { "PotentialQualityIssues": [ "string" ], "QualityScore": number } }, "Customer": { "Audio": { "PotentialQualityIssues": [ "string" ], "QualityScore": number } } },

QualityMetrics is also a sub-section of the QualityMetrics object that you receive through Kinesis CTR events.

QualityMetrics is not available by using the Amazon Connect admin website to view the contact record.

QualityMetrics is not part of EventBridge events.

Symptoms of call quality issues

Following is a list of common symptoms that indicate call quality issues on a participant media connection. You can observe these symptoms on a Amazon Connect call recording on the participant channel.

  • Choppy/broken audio

    • Observation: The audio stream gets interrupted and sounds choppy or broken on the media connection as heard by the listener on the other end.

    • Potential causes: Potentially caused by packet loss due to poor network connectivity leading to a participant sounding choppy or broken to the other party.

  • Delayed audio

    • Observation: Participant experiences a delayed audio from the other side. An effect of delayed audio is consistent overlapping in conversation between the caller and the agent.

    • Potential causes: This can be caused due to constrained network bandwidth/hardware/workstation congestion.

  • Echo

    • Observation: Echo is when agent hears their own voice repeated back to them with a delay.

    • Potential causes: This is often due to audio feedback between the microphone and speaker.

  • Background noise

    • Observation: Extraneous background noise like fans, typing, or call center noise can make it difficult to hear the caller clearly.

  • Distorted audio

    • Observation: Distorted, garbled, or robotic-sounding audio heard by one party coming from the other end.

    • Potential causes: This is typically a sign of a bandwidth issue or faulty hardware.

Analyze the impact on your agents and calls

We recommend using QualityMetrics data together with other fields on the contact record—such as AgentHierarchyGroup and DeviceInfo—to identify the impacted population and spot any trends. Use this information to answer the following questions to understand overall impact:

  • What percentage of agents or calls are impacted?

    • Scenario 1: If only 1 agent is encountering the problem, it could be related to agent workstation including operating system and browser/network configuration of the agent.

    • Scenario 2: If multiple agents in the same hierarchy (for example, the same geographical location or office) experience audio quality issues, this could be a result of a local network issue (modem/ISP/Router/LAN connections) or recent software upgrades to the machines of agents.

    • Scenario 3: Multiple agents (working remote and/or at office location) may be experiencing the issue. Check the browser/system configurations for any updates along with any network changes that may have occurred at the organizational level.

  • What percentage of calls are affected in a given day and out of how many calls?

  • Is the issue observed on incoming calls, outgoing calls or both?

  • Is an entity forwarding calls to Amazon Connect? If so, does the audio quality issue occur in case of direct dials to Amazon Connect without the call quality issue.

Use QualityMetrics

Amazon Connect provides QualityMetrics in the contact record for each connected call. Use the metrics help you to identify the source of the issue.

QualityMetrics contain the following information:

  • QualityScore: An estimate of overall audio quality using a numerical value.

    • Minimum value: 1.00 (indicates poor quality)

    • Maximum value: 5.00 (indicates high quality)

  • PotentialQualityIssues: For calls with potential issues, PotentialQualityIssues is populated with a list of detected reasons which include HighPacketLoss, HighRoundTripTime, or HighJitterBuffer. An empty list suggests that no audio quality issues were detected.

The following list explains the potential values for PotentialQualityIssues and suggests causes to guide your investigation.

  • HighPacketLoss: When this value occurs for PotentialQualityIssues, it suggests that there is packet loss observed on the outbound audio (egress) stream for the participant.

    • Causes:

      • This can occur in the path the packets traverse the network between the participant and the Amazon Connect endpoint which could be due to a bad/poor network, congestion in network, constrained network bandwidth.

      • It can also occur when there could be other applications on the participant's system that may be causing network bandwidth starvation.

  • HighJitterBuffer: This is the time delay introduced by a buffer built-in the browser to correct the order of audio packets after network traversal. Jitter buffer plays a major role in ensuring that the packets received over network at a device are aligned appropriately to provide the audio without distortion.

    • Causes:

      • If a congestion (network and/or hardware) occurs at the participant's end, the JitterBuffer increases causing audio delays/distorted or choppy audio.

      • Jitter buffer is responsible for delaying the processing media packets just enough to smooth out delivery times but high jitter buffer can cause background noise or audio quality issue.

      • If jitter buffer is more than 30 ms or changing very frequently then it means network congestion or low network bandwidth of router. High jitter can also be caused due to hardware issues of the devices involved.

  • HighRoundTripTime: This is the time it takes for a packet to travel through an IP network, from a sending endpoint to a receiving endpoint and back, not including the processing time at its destination. High RTT results in callers experiencing noticeable delays (speech overlap) on the call. RoundTripTime (RTT) is the estimated network round trip time between the participant's device and Amazon Connect endpoint.

    • Causes:

      • The most common cause of high round trip time is a low-bandwidth or constrained network.

      • You may also experience high round-trip time if a certain software program is causing a spike in round-trip time. In the past, some of our customers have reported VPN applications being cause of the issue.

      • If the agent's physical location is distant from the AWS Region of the Amazon Connect instance, it results in an increase to RoundTripTime adding latency.

      • Routing audio through a virtualized desktop (as opposed to redirecting the WebRTC session directly to agent workstation) could also introduce high round trip time.

Next steps

After you have identified whether the issue is HighPacketLoss, HighJitterBuffer, HighRoundTripTime, use the information to troubleshoot the network or the agent's device. See the following topics: