Erlang B vs Erlang C in VICIdial Forecasting

VICIdial's forecasting reports can run two different traffic models, Erlang B and Erlang C. They share the same Erlang input but answer different questions, and picking the wrong one gives you a staffing plan that misses. This post explains the difference in plain terms and when each fits a dialer.

Both models live in the forecasting area covered by the VICIdial reports guide, so start there if the menu is unfamiliar.

The core difference

The split is about what happens to a call that cannot be answered right away. Erlang B assumes blocked calls are cleared: if no line or agent is free, the call is lost and gone. Erlang C assumes blocked calls are queued: the caller waits in a Call queue until someone is free. Same arriving traffic, two different assumptions about the calls that do not get an immediate answer, and that single assumption changes every number downstream.

• Erlang B answers: what is the probability a call is dropped? That is your GoS.
• Erlang C answers: what is the probability a call is queued, and how long will it wait? That is Queue Prob and Average Answer.

flowchart TD
  A[Call arrives] --> B{Agent free}
  B -->|Yes| C[Answered]
  B -->|No| D{Model}
  D -->|Erlang B| E[Call cleared and lost]
  D -->|Erlang C| F[Call queued and waits]
  F --> G[Answered after wait]

Why the answers differ

Because queued callers eventually get served, Erlang C generally recommends a few more agents for the same traffic than B does, since it is also trying to hold wait time down, not just prevent drops. The C math for wait time and queue probability only holds when the agent count is greater than the Erlang load, otherwise the line never catches up and the formulas break down. The average speed of answer that C reports is calculated from the queue probability, the call duration, and the gap between your agent count and the Erlang load, so the wider that gap, the shorter the predicted wait.

When each applies to a dialer

For inbound lines where callers genuinely sit in a queue and hear hold music, Erlang C is the closer fit because it models the waiting. For lines where an unanswered call is simply lost, or for sizing trunk capacity, Erlang B fits better. The forecasting reports default much of the drop-rate math to Erlang B, and the Advanced report adds the C option when you choose report type C.

One outbound wrinkle worth knowing: VICIdial does not log which outbound calls were queued, so on the C-type report the estimated agent figure for outbound work falls back to the outbound drop rate using B-style math. That makes B the more dependable model for outbound sizing, while C earns its keep on inbound queues where callers really do wait and you care about how long.

Picking the right one

Ask what your callers actually experience. If they wait, model the wait with C and watch your Average speed of answer (ASA). If they get a busy outcome and drop, model the loss with B and watch your Abandonment rate against the target Service level. For the daily SLA companion to these forecasts, see the inclusive SLA per day report.

Match the model to the caller experience and your staffing plan stops fighting reality. For a managed VICIdial with both forecasting models built in, see VICIfast pricing.