By CYBERTITO For decades, the Erlang C formula has been the gold standard for contact center workforce management (WFM). It calculates the number of agents required to meet a specific Service Level (such as answering 80% of calls in 20 seconds) based on call arrival rates and average handle times. However, the traditional Erlang C model was built for a single-channel, voice-only era. In today’s modern omnichannel environment—where agents handle concurrent chats, emails, social media messages, and voice calls simultaneously—raw Erlang C leads to severe overstaffing or understaffing. Adapting this classic mathematical model is no longer optional; it is essential for accurate forecasting and operational efficiency.
The Flaws of Traditional Erlang C in Omnichannel Environments
Classical Erlang C assumes a few core conditions: that callers abandon immediately when they don’t reach an agent, that agents handle only one interaction at a time, and that arrivals follow a strict Poisson distribution. In an omnichannel contact center, these assumptions break down completely:
- Multi-Tasking and Concurrency: An agent handling live chat may manage two, three, or even four simultaneous interactions. Traditional Erlang C treats every interaction as a dedicated full-time equivalent (FTE) slot.
- Customer Patience (Abandonment): While voice callers hang up quickly, chat and messaging customers have different patience thresholds. They may step away and return, altering the abandonment curve.
- Blended Workflows: Agents frequently switch between asynchronous channels (like email) and synchronous channels (like voice or live chat), making handling times non-linear.
Modifying Erlang C for Blended and Concurrent Workflows
To make Erlang C viable for modern WFM, analysts and BI developers must introduce correction factors that account for agent concurrency and channel blending.
1. Accounting for Agent Occupancy and Concurrency
When agents handle multiple asynchronous chats, effective occupancy changes drastically. If an agent manages a concurrency factor of $C$, the effective service rate increases. However, human cognitive limits mean that handle times (AHT) often inflate as concurrency increases due to context-switching fatigue. A modified Erlang C formula incorporates a penalty factor ($p$) into the AHT calculation: $$AHT_{effective} = AHT_{base} \times (1 + (Concurrency – 1) \times PenaltyFactor)$$
This adjustment prevents the model from assuming linear efficiency gains that do not exist in practice.
2. Integrating Erlang AB (Abandonment Modeling)
Because digital channels and modern phone queues feature high customer tolerance for waiting (coupled with callback options), Erlang C’s assumption of zero abandonment while waiting distorts requirements. Upgrading to an Erlang AB approach—or introducing an abandonment factor into the waiting queue calculations—results in a much tighter, more realistic staffing band.
Building Automated WFM Pipelines in BI Systems
Modern workforce management relies on dynamic forecasting rather than static Excel sheets. By translating these modified mathematical models into automated SQL scripts or Python pipelines within your Business Intelligence infrastructure, you can calculate staffing requirements in real-time.
Data architects should feed historical arrival distributions, dynamic shrinkage percentages, and real-time concurrency metrics directly into custom WFM dashboards. This ensures that operations teams can adjust interval-level staffing recommendations dynamically, responding to shifts between voice spikes and digital surges without relying on outdated mathematical assumptions.
Conclusion
The Erlang C formula is not dead, but its raw, unadjusted form is obsolete for digital-first operations. By introducing concurrency factors, accounting for cognitive fatigue, and moving toward advanced queuing models, BI developers and WFM professionals can build resilient forecasting engines. Adapting these classic principles ensures that your contact center remains efficient, customer-centric, and accurately staffed across every touchpoint.