Call Center Head-Scratcher: Dynamic Agent Selection
Work-from-home agents pose an interesting problem: how to handle poor quality when calls traverse the Internet.
In talking with one of our customers last month at Enterprise Connect 2018, my colleague Joe LaRosa and I were asked an interesting question about call quality that we couldn't answer.
By way of background, while upgrading company call centers, this fellow was supporting remote, home-based call center agents. Unfortunately, some of the agents were experiencing poor call quality because their call paths traversed the Internet. Internet traffic tends to follow the local time zone, peaking in the morning, around lunch time, when school is over, and then in the evening. Winter and holidays also tend to generate more traffic, impacting the predictable delivery of voice traffic.
The question he posed was whether we knew of a standard way to measure a remote agent's call quality and remove the agent from the queue when the quality dropped below a set threshold. That brought up a pretty interesting conversation, as well as a list of additional questions.
Why Use Remote Agents?
Using agents who work from home instead of at a centrally staffed call center comes with tradeoffs. In general, the costs can be less, primarily due to the lack of a centralized facility requiring power, amenities, and heating and cooling. An additional cost benefit may be obtained through the use of lower-cost, part-time agents instead of full-time staff. Even if you have a central call center, you may find it useful to have remote agents available in order to facilitate continuity of operations should something happen to the central call center. Also possible is using remote agents to handle call-volume surges.
The downside is that remote call agents can be more challenging to manage, especially if your business requires a high level of customer and agent interaction. In addition, developing a good culture can be challenging when agents aren't able to interact with each other and with other staff within the organization.
The conversation that Joe and I had with our customer suggested that his requirement was to provide continuity of operations during the central call centers' renovations.
Agent Call Quality
In our brief conversation, we explored several techniques for measuring call quality.
For one, if a VoIP system supports Real-time Transport Protocol stream monitoring, this could be used to monitor call quality. In this case, the endpoints report call quality every few seconds (typically every five seconds). If a call's quality becomes poor, re-routing the call via the public switched telephone network may be possible. Of course, the call controller and the agent will need the capability of handling the automatic transition of the call.
However, some UC systems only support summary measurements at the end of calls. In such an installation, any action must wait until call termination. In this case, the contact center could remove from the queue an agent whose previous call experienced poor quality until further testing shows that the quality has returned to acceptable levels for the desired duration. It could use synthetic traffic between the call distribution system and the agent to check whether the path has returned to acceptable quality levels.
Alternatively, some voice testing products can continuously monitor call paths, even handling the case where the UC system doesn't support continuous monitoring. These systems have the advantage of being able to detect problems that exist with only one direction, say from the customer to the agent or from the agent to the customer. Over time, any of these systems should be able to collect enough data to determine the time of day that each agent has good call quality, allowing the management system to schedule agents for those times.
We then considered what to do with an agent who has just joined the processing queue. Should the contact center generate a brief call to a test system just to measure and verify call quality? Or should synthetic calls be placed between the endpoints in order to simulate a real call? In both cases, a brief measurement establishes a baseline for comparisons.
Since that brief conversation at EC18, the NetCraftsmen UC team came up with another scenario that we hadn't previously considered. If call volume to the call center increases, should an agent whose voice connection is medium quality be added back into the queue? The premise is that having a mediocre quality call might be better than having an excessively long customer processing queue.
The end result of our discussion was that we weren't aware of a system that provided this level of call center measurement. If you know of one, please leave a comment.