What is queue depth?
Queue depth is the amount of work waiting to be processed at a given moment. It is the accumulated difference between what arrives and what completes. Among operational measures it is the earliest to move when capacity and demand part company.
Why queue depth moves before any completion count does
By the time throughput falls the backlog has been growing for hours; the first person to notice is usually a customer. Queue depth moves earlier because it accumulates every mismatch between arrival and completion, however small. An operations team watching completion counts alone is reading a measure that looks healthy while work piles up behind it. The rate of change matters more than the absolute number, because a steady depth demands a different response from a climbing one.
- Depth accumulates every difference between arrival rate and completion rate.
- A rising depth is visible long before a missed commitment appears.
- Depth by work type locates which category of work is starving.
- An agent fleet scaled to average arrival cannot clear a morning peak.
How queue depth works
- Work is placed on an explicit queue rather than passed directly between steps.
- Depth is sampled at a fixed interval and retained as a series.
- Age of the oldest waiting item is recorded alongside the count.
- Depth is reported by work type so that one category cannot hide another.
- Thresholds are set from the commitment the queue has to meet.
- A sustained rise triggers either additional capacity or a decision to shed work.
How queue depth is evaluated
Queue depth is interpreted against arrival patterns rather than as a single figure. Watch the following over a period that includes a peak:
- depth at the daily and weekly peak
- age of the oldest item in each queue
- rate of change in depth per hour
- depth by work type against its commitment
- time to clear the backlog after a stoppage
Limitations, risks and common failure modes
- Depth is averaged over a day, which conceals the peak that caused the breach.
- Repeated attempts arrive on the same queue, so depth overstates new demand.
- A dead letter queue grows unwatched because nothing measures it.
- Alerts are set on an absolute number that a volume change makes meaningless.
- Capacity is added to clear a backlog whose cause remains in place.