Queue Time

A pharmacy may fill prescriptions in four minutes but make people wait fifteen minutes before a pharmacist starts, so measuring only the four-minute task hides the customer's main delay. Queue time reveals the waiting stage.

Start with a clear unit, because a person waiting at a counter, a customer call waiting for an agent and an approved order waiting for packing each have a different queue, and they should not be combined into one average unless they share a meaningful process. Define the timestamps.

For a call, queue time may start when the caller enters the agent queue and end when an agent answers, so an automated greeting before that point could be included in customer-perceived wait but omitted from a system report. An order can have several waits, such as before credit approval, picking and dispatch, so record each stage if the goal is to find the bottleneck and record total elapsed time if the customer cares about final delivery.

Average queue time equals the sum of measured waits divided by the number of included items, so if 100 customers spend a combined 1,200 minutes waiting, the average is 12 minutes. State whether abandonments and exceptionally long waits are included, because an average can conceal a long tail.

If most people are served in two minutes but a small group waits forty minutes, track a percentile, such as the 90th, as well as the mean if fairness or service commitments matter. Waiting often grows as a process approaches capacity, because a small rise in arrivals can create a large backlog when every available worker is already busy, so adding one employee at the right hour may help more than spreading extra hours evenly across a week.

Demand patterns matter too, since lunch-hour walk-ins or a monthly invoice run can create predictable peaks, and staffing and appointment slots should match actual demand without assuming the arrival pattern will stay the same after a change. Little's Law relates average items in a stable system to arrival rate and average time in the system, which can help check whether work in progress and elapsed time are consistent, but it does not replace direct queue measurement and its boundary must match the waiting stage being analysed.

NIST's discussion of manufacturing flow time treats waiting as a potential form of waste, since an order that is not actively handled can still tie up stock, customer cash or future capacity. Reducing avoidable waits can improve flow without asking employees to rush the actual task.

Look for the cause before adding staff, because orders may wait as approvals are batched once daily, stock locations are unclear or a machine often breaks down, and a different release rule or layout might remove the queue more cheaply. Count customers who leave before service, since a call abandoned after ten minutes is not a zero-minute wait and should not silently disappear from the report, and the same applies to a shopper who leaves a physical line.

Avoid fixing queue time by weakening quality, as a rushed pharmacist may clear a line faster while raising error risk, so pair speed measures with error, satisfaction or rework indicators. For a manager, queue time tells you where customers or work are stuck before service, so measure the correct boundary and the people who leave, then change demand, capacity or process design where the evidence points.