Why Is Packet Loss an Important WebRTC Metrics for Contact Centers?

By Allie Mellen on February 1, 2019
read

Packet loss is an important metric for call centers looking to have high quality audio. Depending upon the level of packet loss, poor call quality can produces a poor customer experience. What is packet loss and why should WebRTC-enabled contact centers work hard to keep it at zero? Read on to find out.


If you’re interested in learning more about the importance of WebRTC in the call center, check out our white paper: Cloud-based Contact Centers: the WebRTC Story.

What is Packet Loss?

Packet loss is just as it sounds: transmitted packets have been lost somewhere along their journey across the network and do not arrive at the receiving endpoint. Packet loss can be caused by any number of things: In wireless networks, inadequate signal strength, interference and poor network hand-offs cause loss; In all network types, excessive congestion, system noise, hardware failures and software corruption can cause loss. Often, one of these failures can have a ripple effect, making things even worse.


There are multiple packet loss metrics that you should monitor to ensure users receive good audio quality. The raw packet loss metric only tells you how much of the transmitted data is lost in transit. It doesn’t tell you whether this lost data is affecting user experiences. To do this, you also need to evaluate post-repair packet loss and discarded packets, which are described in the next section.

What Happens When Packets are Lost?

Fortunately for contact center managers, WebRTC was designed to deliver good quality even when sessions encounter loss, as is typical on public internet networks. Two methods are used to recover from packet loss: Forward Error Correction and retransmission.


The Opus CODEC that is part of the WebRTC specification features forward error correction (FEC) mechanisms that can recover some or all of the information in the lost packet. FEC protects against a single lost packet by sending redundant information to the receiver in separate packets. Referred to as an XOR protection scheme, it enables the receiver to recreate the lost packet.


Sudden bursts of multiple lost packets can be protected by alternate FEC schemes. For example, the sender may transmit multiple FEC packets protecting all packets in a 10 packet window, or all even packets in the window. FEC can adjust and send more redundant data to protect more packets as loss levels change during a session. When loss rates are excessive, the mechanism may be ineffective. In this case, the lost packets are counted as “post-repair packet loss” and audio may be noticeably impaired.


FEC is complemented by concealment mechanisms that help to minimize the audible effects of packet loss. If there is residual loss after repairing packets the receiving system uses error concealment which may result in a slight degradation in audio quality.


Normally, FEC adds 33% bandwidth overhead, which is a worthy trade-off for the additional audio quality it provides. When loss rates spike and FEC has to adjust to protect more packets, the bandwidth overhead also increases.


Retransmissions take advantage of the sequence numbers embedded in each media stream. When the receiver detects a packet is missing from the sequence, it notifies the sender. The sender determines whether to retransmit the packet or adjust the FEC scheme. Time is of the essence for retransmissions because the replacement packet must arrive at the receiving endpoint before the decoder assembles the packets into audio. The sender evaluates the round trip time and determines whether to retransmit. If the retransmitted packet arrives too late, the receiver discards it, which is counted as a “discarded packet.”

 

Why Is Packet Loss an Important WebRTC Metric for Contact Centers?

Ideally, packet loss should be maintained near zero for the best network performance and audio quality. Low loss rates mean FEC and retransmissions consume low bandwidth overhead and the network is running efficiently. When packet loss rates become significant, you should also examine post-repair packet loss and discarded packet metrics. It’s the combination of these three loss metrics that can indicate call quality problems that are noticeable to the users.


High post-repair loss rates and discarded packets can be a huge issue for contact center call quality. It causes WebRTC sessions to send more redundant data, which increases bandwidth consumption. It can also result in audible impairments to the session. Imagine if your agent was communicating a crucial point to a customer, and suddenly that point was gone. Moreover, it can cause customer frustration, as audio quality can be poor. Nothing annoys a customer more than when they cannot communicate their frustrations.


If you’re concerned about maintaining high quality conversations with your customers, and ensuring they have a consistent, high quality experience, you need to be monitoring the packet loss metrics trio.

 

Check the quality of your network with our smart connectivity test. 

Try the Smart Connectivity Test

Tags: WebRTC Use Cases, Amazon Connect, Contact Centers, WebRTC Verticals, WebRTC