Glossary

Every term the rest of the documentation uses, in plain language, with the everyday example first. Nothing here assumes you have done this before.

If a page uses a word you do not recognise, it is defined here and the page should have linked to it. If it did not, that is a bug in the page.

The basic picture

Discrete-event simulation : A way of modelling something by jumping from one moment when something happens to the next, instead of ticking forward second by second. If a car pulls up at 9:00 and leaves the pump at 9:04, the simulation jumps straight from 9:00 to 9:04 — nothing happened in between, so nothing needed computing. That is why a year of a factory can be simulated in seconds.

Entity : The thing moving through your model. A car at a petrol station, a shopper in a supermarket, a customer order in a workshop, a batch of wafers in a chip plant. In the document this is what a release creates and a product describes the path of.

Resource : Something an entity has to use, and which can only serve so many at once. A petrol pump, a checkout till, a milling machine, an oven. See Resources.

Server : One individual unit inside a resource. A resource called tills with "servers": 3 is one queue feeding three tills, not three separate resources. This distinction changes the answer a lot, and the grocery checkout example exists to show by how much.

Queue : The entities waiting for a resource because every server is busy. Two different numbers get called "the queue" and they are not the same thing:

• Queue length — how many are waiting, on average.
• Waiting time — how long each one waits, on average.

They move together but not proportionally, because a busier resource also has more arrivals. A queue that gets five times longer may only make each customer wait four times as long.

Step : One operation on an entity's route: which resource it needs, for how long. A product is an ordered list of steps. See Products and steps.

Route : The full ordered list of steps a product goes through, start to finish.

Words for "how long" and "how much"

Cycle time : How long an entity takes to get all the way through, from arrival to departure — including all the time it spent waiting. This is the number a customer feels. Also called flow time or throughput time elsewhere; this documentation always says cycle time.

Process time : How long an operation actually takes when the entity finally gets the machine. Cycle time minus waiting.

Flow factor : Cycle time divided by process time. A flow factor of 4 means an entity spent three quarters of its life in your plant waiting rather than being worked on. It is the single most useful summary of whether a plant is congested.

Throughput : How many entities finish per unit of time. Cars served per hour, orders shipped per day.

Utilisation : The share of time a resource spends actually working. A pump at 68% utilisation is pumping fuel 68% of the time it is available.

warning

Utilisation is the most reliably ambiguous word in this field, because there are two definitions and they differ by exactly the downtime. clockwright reports busy over available — the definition a capacity decision needs. Some other tools report busy over total, which is a smaller number for the same machine. When comparing against another tool's figure, check which one it means before concluding anything.

Availability : The share of time a resource was able to work — not broken down, not in maintenance.

Work in progress (WIP) : How many entities are inside the system at once, waiting or being worked on.

Little's Law : WIP = throughput x cycle time. It is not an approximation or a modelling assumption — it is arithmetic, true of any system in steady state. It is the cheapest sanity check there is: if your three numbers do not satisfy it, one of them is wrong.

Bottleneck (also constraint) : The resource that limits the whole system's output. Usually, but not always, the busiest one. The petrol station example shows a case where the bottleneck is the smallest resource and nobody would have guessed it from the equipment list.

Words for randomness

Distribution : A description of how a random duration varies. "Four minutes" is a single number; "on average four minutes, sometimes two, sometimes nine" is a distribution. See Distributions, which has a graph of every one.

Mean : The average.

Coefficient of variation (cv) : How spread out a distribution is, relative to its mean. cv = 0 means every draw is identical. cv = 1 is the spread of an exponential distribution, which is the usual benchmark for "quite variable". cv is the number that decides how much queueing you get: queue length scales roughly with (1 + cv²) / 2, so halving variability nearly halves the queue without buying a single machine.

Interarrival time : The gap between one arrival and the next. A mean interarrival time of 90 seconds means 40 arrivals an hour on average — but not one exactly every 90 seconds, unless you say so.

Exponential distribution : The standard model for "arrivals that are independent of each other" — customers turning up without coordinating. Its cv is 1. Most gaps are short and a few are very long.

Erlang distribution : A distribution with a shape parameter k that controls variability: k = 1 is exponential (cv 1), and larger k is steadily tighter (k = 8 gives cv 0.354). Useful for service times, which are usually far less variable than arrivals.

Words for running a model

Horizon : How long to simulate, after the warm-up.

Warm-up : An opening stretch of simulated time whose statistics are thrown away. A model starts empty, which is not how a real plant is at 9am, so the first stretch is unrepresentatively calm. Measuring it would flatter the answer. See Releases.

Steady state : The condition a model settles into once the warm-up transient has passed and the statistics stop drifting.

Replication : One complete run of the model with a different set of random draws. One replication is an anecdote; several give you an idea of how much the answer wobbles.

Seed : The number that determines the random draws. The same seed gives the same run, exactly, every time — which is what makes a result checkable. See Determinism.

Confidence interval : A range around an estimate, with a stated confidence that the true value lies inside. Reported because a single number carries no indication of how much to trust it.

KPI : Key performance indicator — any number you are steering by. Cycle time, throughput, utilisation, the queue at a particular resource. See Reading the results.

Dispatch: who goes next

Dispatch rule : How a resource chooses which waiting entity to serve next, when more than one is waiting. See Dispatch rules.

FIFO (first in, first out) : Serve in arrival order. The fairest rule, the default, and what a single queue at a bank does. Also called FCFS, first come first served.

LIFO (last in, first out) : Serve the most recent arrival first. Rarely chosen deliberately; it happens by accident when work is stacked in a pile and taken off the top. It leaves the average wait unchanged but makes the worst wait far worse.

Priority : Serve higher-priority entities first. A rush order jumping the queue.

Changeover (also setup) : Time lost switching a resource from one kind of work to another — changing the paint colour, swapping a tool, cleaning a vessel. A resource doing changeovers is neither working nor idle. See Dispatch rules for grouping work to avoid them; the paint shop example is entirely about this.

Campaign : A run of consecutive jobs of the same kind on one resource, to avoid changing over between them.

Interruptions

Downtime : Any time a resource cannot work. See Downtime.

Breakdown : Unplanned downtime, arriving at random.

MTTF (mean time to failure) : The average working time between one repair finishing and the next failure.

MTTR (mean time to repair) : The average length of a repair.

MTBF (mean time between failures) : MTTF + MTTR. Named here only because other tools use it and the two are easy to confuse — clockwright asks for mttf_seconds, the working time, not the whole cycle.

Preventive maintenance (PM) : Planned downtime — scheduled servicing, either on a calendar or after so many pieces.

Batching and materials

Batch : A group of entities processed together as one load. An oven that fires 30 panels at a time is batching. See Batching.

Consumable : Something used up rather than occupied — fuel, adhesive, packaging, a chemical bath. A resource is borrowed; a consumable is spent. See Consumables.

Stockout : A moment when an entity needed a consumable and there was not enough, so it had to wait.

Reorder point : The stock level at which a replenishment order is placed. The petrol station example measures what happens as you lower it.

Lead time : How long a replenishment takes to arrive after being ordered. The delivery truck's journey.

Operator : A person needed in addition to a machine, drawn from a shared pool. A resource can be free and still unable to start because nobody is available to load it. See Operators.

Things that go wrong

Scrap : An entity destroyed part-way through and never completing.

Rework : An entity sent back to repeat a step.

Transport : Time spent moving between resources. It occupies the entity but not any machine, so it adds to cycle time without adding to anybody's utilisation.

Starvation : A resource sitting idle because nothing has arrived for it to work on.

Blocking : A resource that has finished a job but cannot release it, because wherever it goes next is full — so it cannot start the next job either.