Remote Contact Centre Agents: Voice Quality Over VPN

Remote Contact Centre Agents: Voice Quality Over VPN

Moving a contact centre team to home working changes the voice path for every conversation. An office agent typically sits a few switches from the phone system and reaches the internet through a single controlled egress. A home agent connects over an uncontrolled access network, often through a consumer router, a wireless link, and a VPN tunnel whose configuration nobody in the organisation chose. The audio still arrives, which is why the problem is easy to miss in a functional test and hard to live with in daily use: calls connect, agents work, and quality problems get attributed to the agent rather than to the network. Designing for remote voice means accepting that the last mile is beyond your control and concentrating effort on the parts that are not: prioritisation, measurement, endpoint behaviour, and giving agents a way to report quality problems with useful evidence.

What the Home Network Adds to the Path

The office path is typically a handful of well-behaved hops with consistent queuing. The home path adds a residential access segment that may be congested, a Wi-Fi link that may be weak or share airtime with a household video stream, a consumer router with its own buffer behaviour, and a VPN tunnel that adds encryption overhead and constrains the path. Each addition contributes delay, jitter, or loss, and voice quality degrades non-linearly as these accumulate.

The Tunnel Question

Whether voice traffic traverses a VPN is a genuine design decision rather than a default. A full-tunnel VPN gives control over the path and consistent routing but adds a dependency on the tunnel, and traffic competes with whatever else the agent is doing on the same tunnel. Split tunnelling for voice reduces contention on the tunnel but can produce asymmetric paths, where outbound and inbound traffic take different routes, which is a common cause of one-way audio that is very hard to trace.

Whichever approach is chosen, voice should be treated as a priority class on whatever link the agent actually has. If the same access link carries file synchronisation, cloud storage traffic, and video conferencing alongside the call, contention is likely. Some organisations reduce contention simply by discouraging concurrent high-bandwidth use during shift hours, which is a policy measure rather than a technical one, and often more effective than a policy that assumes the technology will resolve it.

Prioritising Voice Across the Remote Path

• Classify real-time media as the highest-priority traffic on any link under your control
• Verify that classification is preserved from the agent's endpoint through to the phone system
• Ensure the VPN or WAN device does not re-classify inbound marks unexpectedly
• Size tunnel and bandwidth for concurrent sessions rather than per-session figures
• Review queuing behaviour under load, since an idle link does not reveal queue problems
• Keep firmware and configuration of the agent's softphone and driver current, since defects there look like network faults

Measurement: Giving Agents Useful Evidence

Agents cannot diagnose their own network, but they can collect evidence that makes diagnosis possible. A quality indicator drawn from the softphone — round-trip delay, jitter, packet loss percentage, and the codec in use — attached to a report of when and where the problem occurred, turns an unusable complaint into a workable case. Without this data, the support path defaults to resetting the endpoint, which sometimes helps and frequently does not, while the underlying condition persists.

The measurement should be designed so that the agent can produce it in under a minute during a live call, and so that the data lands somewhere the support team can aggregate. Aggregating this across the team also reveals patterns that individual reports miss: whether a problem is specific to one agent's setup, specific to a region, or specific to a time of day when the access network is congested.

Endpoint Practices That Matter More Than Expected

• Use a wired connection where the home setup allows it, and treat Wi-Fi as the fallback
• Position the endpoint and any attached audio device away from household noise sources
• Use a headset rather than laptop speakers, since speaker pickup causes echo for the far end
• Avoid USB bus power savings and aggressive sleep behaviour on the endpoint device
• Keep the softphone and audio driver at tested versions, and document the known-good combination
• Provide a second approved headset or device so an agent can isolate an equipment fault quickly

Redstone Approach

Redstone can support remote agent deployments by advising on how remote endpoints should be configured to register with Redstone platforms, and how voice traffic should be classified along the path to the call centre platform. Redstone can also help structure the call handling side — how queue and agent groups are organised, how supervisors escalate to remote staff, and what reporting is available to see queue and call performance for a distributed team.

Because the home network cannot be engineered, Redstone's emphasis is on making the controllable parts correct and on giving agents a clear escalation path with evidence. Redstone can advise on a reference endpoint setup and on the test plan used to validate a new agent's connection before they take live calls.