The reason a queue can move faster with fewer servers

A bank has three tellers, each with a separate queue. A consultant proposes merging the three into one queue feeding all three tellers. Average wait time drops, even though the number of tellers and the arrival rate are unchanged.

What everyone sees

Three separate queues look fair and fast: customers pick the shortest one, and three lines should move faster than one. The possibility that choice itself is the problem — that selecting among lines introduces variance that a single queue eliminates — is counterintuitive.

What is actually happening

Queueing theory shows that a single queue served by multiple servers (an M/M/c system) has lower average waiting time and lower variance than multiple single-server queues (c independent M/M/1 systems) at the same total service rate. The reason is pooling: a single queue ensures that no server idles while customers wait in another line. Larson described the slowed-queue illusion: customers perceive the line they are not in as faster, which compounds dissatisfaction in multiple-queue systems. Maister reviewed the psychology of waiting and found that perceived fairness (first come, first served) reduces the subjective cost of waiting beyond the objective time saved.

Why it stays hidden

The improvement hides because separate queues feel like more capacity. Three doors look faster than one, and the suggestion to remove two doors looks like a reduction. The gain comes from eliminating the mismatch between queue choice and server availability, which is a coordination failure that the customer cannot see from their position in the line.