Learn how to calculate base staff for randomly arriving workload, and the tradeoffs between the widely-used queuing formula, Erlang C, and alternatives to it. Also learn how to calculate staff for deferred work.
- How many people do you need to handle the customer workload on a Monday morning at 10 a.m.? How about a Thursday afternoon at 2 p.m.? Let's look at steps four and five of the planning process to find out. Here I'll introduce the methods used to accurately predict staffing levels. We'll look at the popular queuing formula Erlang C, discuss an important alternative to it, and we'll see how staff and system resources are related to each other and why they must be considered together. To determine staff requirements for randomly arriving work that needs to be handled as it occurs, many contact centers use a formula called Erlang C.
Erlang C's used worldwide. It's trusted for its accuracy, but as with any formula, it's not perfect. Erlang C assumes that you'll handle every contact so it can over estimate requirements if some of those customers abandon before reaching an agent. Now that's not as big of a problem as it might initially seem. You got to remember that too few staff are what lead to long queues and abandonment, so the issue largely takes care of itself in real world application. Here's a look at the formula. Now fortunately, we don't have to use it in its raw form.
Erlang C software calculators are readily available from many sources and the formula's built into workforce management systems. This data is from and ICMI staffing calculator called Queue View, which you can access through the links in the exercise files. This and other Erlang C programs require you to input four variable. The first is average talk time. Let's say the talk time for the increment you're looking at is 180 seconds, three minutes. Next is after call work.
Let's say you have about a half a minute, 30 seconds, so you plug in 30. Then you'll enter the number of contacts you forecast for the half hour. And finally, you'll enter your service level objective in seconds. If your objective is to answer 80% of calls in 20 seconds, you plug in 20. If your objective is 90% answer in 15 seconds, you'd plug in 15 and so forth. Voila, you get a table that illustrates base staff requirements. The first row shows service level at 24% of calls being answered in 20 seconds.
Now, we're thinking to ourselves, that's too low. So, let's see. 45% in the next row, nope still too low. 61%, 73%, 82%, well let's say our objective is 80% in 20 seconds, so that row 82% is closest. Now, I always recommend going with the row at or above just to be safe. As you can see in the first column, you need 34 agents that half hour, not 30, not 35.
This is precise. Average speed of answer shown in the next column will be just under 13 seconds. Occupancy will be 86%, and that's the time those 34 agents will be handling calls, versus waiting for them to arrive, and in the last column, trunkload, is the traffic on the network converted to and expressed in hours, which will make sense to your telecom and IT team. They can then use these figures to determine the minimum system capacity that you need. If you're under staffed and service level drops, you'll need more system capacity because callers are waiting in queue and tying up resources.
That's why staffing and system resources must be considered together. As you can see, when you have a good estimate of the workload for an agent group, down to a specific interval, calculating agents needed is straightforward. Just remember, ideally you need to do this for every interval as you look out into the future, and for every agent group handling different contacts. That's where workforce management software, if you have it, can be a big help. With chats, some social interactions or other types of work that have back and forth, your agents might be able to handle more than one customer at a time.
In that case, you can divide base staff requirements by two or even three to represent concurrent sessions, but I always recommend playing it safe. You don't want to overwhelm your agents or see quality or service levels suffer. You can make adjustments as you go along as you see how thing are unfolding. What if you have skills-based routing where you need calls routed to agents with those skill sets, say around languages or specific knowledge requirements. Erlang C assumes all agents in a group can handle the contacts presented to that group, that they're sharing work.
To a degree that's not the case, you may want to experiment with computer simulation, an alternative to Erlang C. Like a flight simulator, staffing simulators don't tell you what to do, they tell you what happens when you try different variables and combinations. You can then begin to zero in on the resources you need. There are a number of simulators on the market, and some workforce management systems provide simulation modules that are built right in. Finally, let's turn to workload that can be deferred, such as email or contacts whose timing you determine, including outbound calls.
Here you'll use a traditional units of output approach. You have this many widgets to produce, so you need that many people. The response time formula requires you to input three variables: volume, average handling time, and your response time objective. So here's an example, if you have 200 email messages you need to handle within two hours or 120 minutes, and they take let's say an average of four minutes, you'll know you need a minimum of seven agents for that work within that time frame or 200 over 120 divided by four.
The main message in this video is don't guess at staffing. Use proven methods to be as specific and accurate as possible.
- Recognizing key contact center trends
- Realizing the value created by contact centers
- Developing a customer access strategy
- Identifying performance gaps
- Choosing service level and response time objectives
- Forecasting your workload
- Establishing a long-term staffing plan
- Forming measurable objectives
- Monitoring and coaching individuals