The Kingman formula is a factory-physics equation that estimates how long jobs wait in a queue before being processed, showing that queue time explodes as a workstation approaches full utilization and as variability rises. Also called the VUT equation, it is one of the most important results in queueing theory for manufacturing because it explains why a line that looks "efficient" on paper can still deliver painfully long and unpredictable lead times. Understanding it changes how you think about running machines hot, batching work, and taming variation.
Named after British mathematician Sir John Kingman, the formula approximates the average waiting time in a single-server queue where arrivals and service times are variable (the G/G/1 case). In its manufacturing-friendly form it is written as three factors multiplied together:
Wait time in queue = V × U × T, where:
The V factor is usually expressed as the average of two squared coefficients of variation: one for the gaps between arrivals and one for the process time. The U factor, written as ρ / (1 - ρ), is where the drama lives. As utilization ρ climbs toward 1.0, the denominator shrinks toward zero and the whole expression rockets upward. This nonlinear blow-up is the single most counterintuitive lesson of factory physics.
Managers love high utilization because idle machines feel like wasted money. But the Kingman formula shows that the relationship between utilization and waiting is not linear, it is hyperbolic. Push a machine from 80 to 90 percent utilization and the U factor roughly doubles. Push from 90 to 95 percent and it doubles again. The last few points of "efficiency" cost enormous amounts of queue time, work in process, and delay.
This is why chasing raw machine utilization as a target often backfires. A workstation deliberately run with slack can move jobs through faster and more predictably than one pinned at the red line. If you are trying to raise throughput without understanding queueing, read our primer on throughput in manufacturing and how it differs from utilization, plus our guide to capacity utilization and the traps that come with it.
Consider a CNC cell. Jobs arrive on average every 12 minutes, and each takes 10 minutes to machine. Utilization is therefore 10 / 12, or about 0.833. Suppose the coefficient of variation for arrivals is 1.0 and for process time is 0.75.
Multiply them: 0.781 × 4.99 × 10 = about 39 minutes of average queue time. Total time in the cell is roughly 39 + 10 = 49 minutes.
Now raise utilization to 0.95 by feeding more work in, keeping variability the same. The U factor jumps to 0.95 / 0.05 = 19. Queue time becomes 0.781 × 19 × 10 = about 148 minutes. A modest 12-point increase in utilization nearly quadrupled the wait. That extra WIP piling up in front of the machine is a direct, predictable consequence of the math.
The Kingman formula makes it clear you have two knobs, not one. If you cannot afford to lower utilization, you can attack the V term instead. Halving your combined variability roughly halves queue time at any given utilization. Sources of process variability on the floor include unplanned downtime, setup and changeover swings, quality rework loops, and inconsistent operator methods.
Practical variability-reduction moves include:
Kingman tells you the wait; Little's Law then converts that wait into inventory. Little's Law states that WIP equals throughput multiplied by flow time, so any queue time the Kingman formula predicts translates directly into parts sitting on the floor tying up cash and space. The two equations together form the backbone of factory physics.
The formula also reinforces the theory of constraints. Your bottleneck is usually the workstation running closest to full utilization, which is precisely where the U factor is most explosive. Protecting that resource with a buffer, as in a drum-buffer-rope scheme, absorbs upstream variability before it reaches the constraint and keeps queue time from spiraling.
The formula is an approximation, not a laser-precise predictor, and it assumes a single server without balking or priority rules. But its directional lessons are ironclad: never target 100 percent utilization on a shared resource, treat variability as a first-class enemy, and expect lead time to be nonlinear in load. When someone proposes squeezing "just a bit more" out of a machine already at 90 percent, the VUT equation is your evidence for what it will cost in delay and WIP.
The Kingman formula only helps if you can measure the inputs, and most factories cannot see their real utilization or variability with any accuracy. Fabrico is the real-time data foundation that closes that gap. Our real-time OEE and production monitoring captures actual cycle times, stoppages, and utilization at each workstation, so the U and V terms in your queueing analysis come from live shop-floor data rather than guesswork. Fabrico's computer vision reads machines that have no PLC, which means even legacy cells become visible.
Because variability is so often driven by unplanned downtime, Fabrico's field-ready CMMS gives you work orders, asset records, preventive scheduling, and spare-parts tracking to attack the maintenance-driven variance head on. If you are new to that side, start with our explainer on what a CMMS is and how it supports overall equipment effectiveness. Fabrico is EU-built with EU data residency, so your production data stays in Europe.
It is an approximation for the general single-server (G/G/1) queue and is most accurate when the workstation is running at moderate to high utilization. It assumes stable, well-behaved arrival and service processes without balking or complex priority rules. For rough capacity planning and for understanding directional trade-offs it is extremely reliable, but for precise scheduling you would validate against measured data or a discrete-event simulation.
There is no universal number, but the formula shows the marginal cost of load rises steeply above roughly 85 to 90 percent on a shared resource. Many high-mix shops deliberately hold non-bottleneck stations well below full load and reserve near-capacity operation only for the constraint, which is buffered separately. The right target depends on how much variability you carry: the noisier your process, the more slack you need to keep lead time sane.
They answer different questions. Kingman predicts how long a job waits in queue as a function of variability, utilization, and process time. Little's Law relates the average WIP, throughput, and flow time in any stable system. In practice you use Kingman to estimate the waiting time, then apply Little's Law to translate that waiting into how much inventory will accumulate.
Want to see your real utilization and variability instead of estimating them? Book a Fabrico demo and watch live OEE, cycle-time, and downtime data turn factory-physics theory into decisions you can act on today.
Login
Blog