Reimagining Staffing: Why Erlang C Is No Longer Sufficient

In GWFM’s latest learning, we research the differences between #ErlangC and #ErlangX —a modern approach to workforce planning that helps contact centers overcome inefficiencies, reduce costs, and confidently meet customer demands.

Could outdated staffing methods be blamed **Reimagine Staffing: Limitations of Erlang C in Contemporary Contact Centers**

Outdated staffing methodologies may contribute to challenges such as understaffing and the failure to achieve service-level objectives.

The traditional Erlang C model has long been considered an industry standard. However, it does not sufficiently address the complexities faced by today’s dynamic contact centers, particularly crucial elements such as customer impatience, call abandonment, and redials.

In GWFM’s latest educational publication, we analyze the differences between Erlang C and Erlang X—a modern approach to workforce planning. This innovative solution aims to assist contact centers in overcoming inefficiencies, reducing operational costs, and effectively meeting customer demands. for understaffing, understaffing, or missed service levels?

As we pointed out before, the traditional model of Erlang C is often regarded as the industry standard. However, it falls short of meeting the demands of today’s dynamic contact centers, failing to address critical real-world factors such as customer impatience, call abandonments, and redials.

In GWFM’s latest learning, we research the differences between #ErlangC and #ErlangX —a modern approach to workforce planning that helps contact centers overcome inefficiencies, reduce costs, and confidently meet customer demands.

Why Workforce Management Needs Erlang Models

Improving and analysing contact centers can be an exciting challenge, as it allows us to apply a wide array of mathematical theories, something we love at CCmath. One key area of focus in this process is Workforce Management, which aims to optimise contact center operations and create accurate capacity planning.

Forecasting: The First Step in WFM

The first step in the Workforce Management cycle is forecasting. This involves analysing historical data, such as incoming call volumes and Average Handling Time (AHT), to predict future volumes. When multiplying call volumes with AHT, we derive the net workload, the amount of agent time needed to handle a certain amount of work without accounting for safety staffing. 

However, forecasting isn’t straightforward. Call volumes follow a Poisson process, making them unpredictable and prone to fluctuation. To address this, smoothing techniques are applied to reduce noise and reveal patterns in call volumes.

For those unfamiliar with WFM terms like AHT or net workload, we have provided a WFM Definitions page where each term is clearly explained. You may find it helpful to review these WFM definitions before diving deeper into the process.

The Challenge

This natural fluctuation makes it difficult to detect clear trends or seasonality. To make accurate predictions, a smoothing process is essential. By applying smoothing techniques, we can reduce the impact of these fluctuations and create more reliable forecasts, leading to better staffing decisions and improved overall contact center performance.

In the graph below, we see an example of a dataset driven by a telephony platform (in blue), where the actual values fluctuate, while the forecast (in orange) remains much smoother. This smoothing helps identify underlying patterns in call volume, which is useful for accurate forecasting. 

However, this smoothing process introduces a significant problem. Since the forecast is smoother than the actuals, it can sometimes be higher or lower than expected. When the forecast is lower than expected, it means more agents are needed than initially anticipated, leaving the contact center understaffed. Therefore, we must account for this issue when generating the forecast and adjust it to include a safety buffer to prevent understaffing. 

To solve this issue, we need a way to close the gap between the smoothed forecast and what’s actually needed. Adding safety helps make sure the contact center has enough staff, even if the forecast misses higher demand, reducing the risk of being understaffed.

Role of Erlang Models in Workforce Planning 

This is where the Erlang models come into play. In the graph below, we see a plot of the actual net workload – call volumes times the average handling time – (blue line) and the two forecasts of the workload: one without a safety margin (orange line) and one with a safety margin (green line). A forecast with safety typically anticipates higher volumes than the actual workload and the forecast without safety, ensuring sufficient staffing to handle unexpected demand. 

This approach traces its roots back over a century to the pioneering work of Danish mathematician Agner Krarup Erlang, who in the 1910s introduced the Erlang C model, a mathematical formula designed to calculate safety staffing levels in call centers.

Erlang C: The Original Workhorse

How Erlang C Works
In the Erlang C model, we assume a system like the one shown in the diagram below. Call arrivals are random and independent, meaning they happen unpredictably but follow a pattern over time. Similarly, the time agents take to handle each call varies, but follows a consistent average. In the example, there are 5 agents (servers) available to take calls. One key assumption in Erlang C is that the waiting line for customers is unlimited, so there is always space for more people to wait if all agents are busy.

The Erlang C formula calculates the probability that an arriving call (or customer) will need to wait before being served. Based on this, we can compute the service level (SL – the percentage of calls answered within the average waiting time) and the average speed of answer (ASA). We can use the Erlang model with the forward or backward computation. In the forward computation, we obtain the performance measures (SL, ASA) based on the available inputs (forecasted call volumes, average waiting time – AWT and average handling time – AHT). The backward computation calculates the required resources (Number of Agents) or feasibility of targets (SL or ASA), based on the desired performance measures. 

Limitations of Erlang C
Erlang C is a useful model, but it has some limitations. Firstly, it assumes infinite patience, meaning that all customers will wait indefinitely, which is not realistic in modern contact centers. Secondly, it assumes that agents handle calls at a constant rate. Thirdly, in the Erlang C model, there are no call abandonments. It assumes that customers do not hang up or abandon the queue, which often leads to an overestimation of the needed agents. This is the reason why Erlang X was invented.

Erlang X: The Modern Solution

Erlang X was developed as an improvement over Erlang C to better reflect realistic customer behaviours, particularly impatience, abandonment, and redials. We define abandonments as customers who leave the queue if they wait too long, and redials as calls that may return to the queue after being abandoned. Unlike Erlang C, which assumes customers either wait indefinitely or call only once, Erlang X accounts for the probability of abandonment, represented by an abandonment rate and incorporates redials, acknowledging that some customers may call the center again.

In this context, agents refer to servers who handle customer calls. The diagram below illustrates a queue with fixed capacity, meaning new callers are blocked from entering when the queue is full. The blue arrow represents customers who abandoned the queue, while the green arrow indicates those who redialled the contact center. Incoming calls originate from fresh customers who have not yet reattempted contact.

How Erlang X Works
The Erlang X model also offers forward and backward computations. Forward computation determines key metrics such as occupancy (the ratio of time an agent spends handling calls compared to their total available time), ASA (Average Speed of Answer), SL (Service Level), and the percentages of abandonment and blocking. It does this by using inputs like the required number of agents, the number of customers waiting in the queue, forecasted call volumes, Average Handle Time (AHT), Average Wait Time (AWT), redials, and customer patience (how long a customer is willing to wait before abandoning or redialling). Backward computation takes the results from the forward computation and works in reverse, using these outputs as inputs to calculate the original parameters needed for forward computation.

Example Comparison
An example to illustrate the differences between Erlang C and X could be the following: In a holiday rush, Erlang C might suggest 50 agents to hit an 80% service level, assuming everyone waits. But Erlang X, accounting for hang-ups, might show only 45 agents are needed since not all customers will wait. This adjustment reflects a more accurate staffing requirement and prevents unnecessary overstaffing, thereby reducing costs.

Erlang C is suitable for simpler scenarios with high customer patience, while Erlang X is better for complex environments where customer abandonment is a factor. The key difference lies in handling waiting times and staffing. Erlang C assumes callers wait in a queue until served if all agents are busy. In contrast, Erlang X accounts for customer abandonment and dynamically adjusts staffing to reduce wait times and balance the load more effectively. Think of Erlang C as a strict teacher, while Erlang X is the cool one who adds extra desks when needed.

In conclusion, whether we are relying on the dependable but structured approach of Erlang C, or the adaptive and modern flair of Erlang X, the key is understanding your service environment and the right balance of staffing. Imagine that busy contact center once again, with every incoming call offering an opportunity to brighten someone’s day or a test of patience. With the right model, you’re not just managing queues—you’re crafting moments of connection.

Why Erlang X Matters Today

This is where Erlang X truly shines, stepping beyond traditional assumptions to deliver smarter, more flexible solutions tailored to the realities of today’s customer interactions. It’s not just about meeting numbers; it’s about creating the kind of efficiency that lets your team greet every call with a cheerful “Hello!”—and keep costs under control while doing so.

Moreover, Erlang X has been expanded to handle chat concurrency with Erlang Chat, recognizing the importance of multi-channel communication in modern contact centers and ensuring optimal staffing across both calls and chats.

Source: GWFM Research & Study