Staffing Your Workforce in an Omni-Channel World 2017-12-15T13:36:41+00:00

Staffing Your Workforce in an Omni-Channel World

By Eric Hagaman, Aspect Software

chat1Human beings are communicating through more channels than were ever dreamed of 20 years ago, and rapid smartphone adoption has put all of these channels in the palms of our hands. With 30% of worldwide mobile users having smartphones, consumers have come to expect ubiquitous communication in the voice channel as well as in newer modes of communication such as chat, SMS, email, social media, and dedicated apps. In this increasingly omni-channel world, the contact center is becoming the defacto clearinghouse for customer  communication, but with new customer interaction channels being required on a significant scale, the old voice workforce models are in need of renovation. Workforce planning professionals must adapt to the unfamiliar and often complex dynamics of conversations in these channels in order to optimize staffing for those most coveted objectives of great customer experience and low cost.

The contact center industry has spent decades refining the science of forecasting, scheduling, and intra-day  tracking of adherence for inbound and outbound voice calls. The next frontier is creating the same level of  understanding for management of the workforce within each and across all of these channels. Unfortunately, these other channels don’t have the same characteristics as the voice channel, and you can’t use the same planning and forecasting techniques to manage the workforce in these non-voice channels.

Let’s consider the chat channel. Forrester notes that chat is the third most heavily used form of interactive customer communication after voice and email, but of  these it is by far the fastest growing at about 8% per year. Unlike the voice channel, it is common practice to assign multiple simultaneous conversations to a single agent in the chat channel. Assigning one chat contact to an agent is generally considered underutilization of resources due to gaps in time spent waiting for customer  message composition after the agent has composed and sent his or her reply. In the simple example below, a single chat channel is assigned to an agent, and the agent does, in fact spend quite a bit of time waiting idly for the customer to compose the next message.

Single Channel Chat

chat2

Now let’s consider two customer chat channels with customer composition times and agent composition times similar to those above. From the agent perspective,  this is quite a different experience.

Two Channel Chat

chat3

As you can see, the agent has significantly less wait time, in fact in several cases he or she is going immediately from composing for one channel to composing for the  other channel. The customers also start to feel the effects with slightly longer waits for the agent’s response.

If we were to add a third chat channel to the mix, the agent would have virtually no wait time, and customers would start to experience noticeably longer delays in  agent response. When an agent works on multiple simultaneous chat sessions, each session’s handle time is increased over the duration that it would be if the agent were chatting with customers one at a time. The average session time of an agent increases with the number of simultaneous sessions in which the agent is engaged. Further, the number of simultaneous sessions in which an agent is engaged increases with the number of chats offered. The unfortunate and inescapable conclusion  is that the average chat session time is mathematically related (in a very complicated way) to the number of chats offered. That is not the case in a traditional voice-only contact center where the AHT for a call is assumed to be independent of the number of calls offered, so you can’t use traditional contact center  Erlang-based thinking for chat.

Without any other guidance, most workforce managers would use some simple rules of thumb to estimate staffing levels required to meet service level targets. Their estimates would likely be rooted in their experience with voice, and here is a specific example to consider on how you could get into trouble trying to staff for chat using traditional voice models. A team of chat agents has been assigned to provide customer service, with a maximum of three concurrent chats per agent. Using the pattern of historical chat volume, the workforce planner observes the average handle time of an employee handling one chat at a time and divides that by 3 (since  each agent is handling up to a max of 3 sessions) then uses standard single skill voice channel Erlang-based methods to arrive

at a required number of agents for  each time period to meet a 70% in 30 seconds service level goal. Using this method, the staffing looks like the Current Industry Practice line on the charts on the next page.

However, when an agent’s workload goes from one chat session to two and then to three, new dynamics come into play, including the following:

  • Agent unproductive time is now impacting multiple customer chats.
  • As mentioned above, from the perspective of a customer, the agent is now responding more slowly because he or she is working on replies to multiple other customers interwoven with their responses to a single customer.
  • The agent is now switching mental and business contexts as his or her attention moves from one customer to another, and some time is spent getting back up to speed on the contents of that conversation before composing a reply. For complex concepts, this time can be significant. When handling three or more simultaneous chats, the agent can easily get confused and have to reread previous messages.

The actual number of chat staff required to meet expected service levels (taking into account all of the dynamics of the chat channel) is shown in the Actual Staff Required line below.

chat5
You can see that the actual number of staff required is much higher than what would be estimated using current industry practice. Without access to a tool that  incorporates the complex mathematics of chat, the workforce planner would have dramatically understaffed the chat team.

Another way to look at this is from the perspective of service levels. In the chart below, the Current Industry Practice line shows that the contact center would be well short of their required service level if they were to schedule staff based on estimating handle time and using voice channel methods. Aspect has developed a new channel modeling tool that gives workforce planners a powerful new weapon in their workforce planning arsenal. Since the mathematics of chat are too complex to solve algebraically, it uses a Monte Carlo simulation technique to make accurate chat staffing forecasts. The Aspect Multi-Chat Calculator, takes into account all of the dynamics unique to chat. As shown below, the actual service level would be very close to being met throughout the day if we staffed using the Multi-Chat Calculator as depicted in the Aspect Model line.

chat6

The industry has had years to think about how to plan the voice workforce. Not so for other customer communication channels. When you start to peel back the onion, you will find a whole new set of metrics that we need to incorporate in our forecasting models such as chat concurrency (how many simultaneous chats is the agent handling), agent composition time, customer composition time, agent wait time, customer wait time, and number of messages exchanged in a session, to name a few. For WFM professionals, it is a radically different way of having to think about customer engagement.

Probably few readers would disagree with me when I assert that the use of non-voice channels will become increasingly important. If the trends of the past few years continue into the future, we will see growth in voice channel volumes, but voice will become a smaller percentage of the total mix of customer interactions. Without new channel modeling techniques, workforce planning for non-voice interactions will become increasingly inaccurate. In the next year or two, expect to see other  workforce management vendors featuring improved forecasting models with non-voice channel capability.

Eric Hagaman is product manager for Aspect Software, focusing on Aspect Workforce Optimization technologies with a particular emphasis on workforce  management. Eric monitors the pulse of the market to identify new trends and approaches to workforce optimization, looking for those product enhancements that will provide the most value to customers and help them master the next generation customer contact.