Dispatching rules are simple priority rules that decide which job a work center runs next when several are waiting: first come first served, shortest processing time, earliest due date, critical ratio. Each optimizes a different goal, so the rule you pick decides what the work center is good at.
Stand at any busy machine and you will see the same moment repeat all shift: the job finishes, and three or four others sit in queue, waiting. Which one goes next? That single decision, made hundreds of times a day across a plant, quietly sets your average lead time, your on-time delivery, and your work-in-process. A dispatching rule is just a written answer to "what next" so the decision is consistent instead of whoever shouts loudest. This post covers the common rules and the KPI each one is built to win.
What are dispatching rules?
Dispatching rules are simple, local priority rules that pick the next job to run at a work center from the jobs currently queued there. They are local because each work center applies its own rule to its own queue, using information it already has, processing time, due date, arrival time, without solving the whole plant's schedule at once. That makes them fast and easy: no optimization engine, no full model, just a rule anyone can apply at the machine. It also makes them limited, because a rule that is smart for one machine can be blind to what happens three operations downstream.
Dispatching rules live one level below the plan. The master production schedule and the broader production schedule decide what to make and roughly when; dispatching rules decide the fine-grained order of execution at each machine in real time as reality shifts. When a plant lacks a formal advanced planning and scheduling engine, dispatching rules are often the entire scheduling logic on the floor, which is why choosing them on purpose matters.
What does each dispatching rule optimize?
Each rule is built to win a different metric, and none wins them all. First come first served (FCFS) runs jobs in arrival order; it optimizes fairness and predictability, not speed, and it is the default when nothing else is specified. Shortest processing time (SPT) runs the quickest job next; it provably minimizes average flow time and work-in-process, clearing the most jobs fastest, at the cost of letting big jobs languish. Earliest due date (EDD) runs the job due soonest; it minimizes the worst lateness, the maximum a single job runs past its due date, which makes it the go-to when hitting dates is the point. Critical ratio (CR) weighs time left against work left; it dynamically balances urgency and workload so the job most at risk of being late goes next.
The critical ratio deserves a closer look because it is the only common rule that changes as time passes. You compute it as the time remaining until the due date divided by the work time remaining on the job. A ratio below 1 means the job is behind and needs to jump the queue; a ratio of exactly 1 means it is right on pace; a ratio above 1 means it has slack. Because both the numerator and denominator move as the clock runs and work gets done, the same job's priority rises automatically as its due date approaches, which is why CR adapts where the static rules do not.
Which rule should a work center use?
The right rule follows the KPI the work center is judged on. If the goal is to blow through a backlog and free up work-in-process, SPT clears jobs fastest. If the goal is hitting customer promise dates, EDD or critical ratio protect due dates. If fairness and simplicity matter more than optimization, FCFS is honest and unarguable. Many plants blend: EDD or CR as the base to protect dates, with SPT as a tiebreaker among jobs that are equally safe on time. Match the rule to what the plant measures on schedule attainment and revisit it when the goal changes.
One caution about SPT: because it always favors the quick job, a stack of long jobs can sit untouched shift after shift while short work keeps jumping ahead. Plants that lean on SPT usually add an aging rule, once a job has waited past a set limit, its priority is bumped, so the throughput win does not come at the cost of a few orders that never run. The lesson holds for every rule: watch not just the metric it optimizes but the jobs it quietly leaves behind.
| Rule | Runs next | Optimizes | Weakness |
|---|---|---|---|
| FCFS | The job that arrived first | Fairness, predictability | Ignores urgency and job size |
| SPT | The shortest job | Average flow time, low WIP | Long jobs can starve; ignores due dates |
| EDD | The job due soonest | Maximum lateness | Ignores how much work each job needs |
| Critical ratio | The lowest time-left / work-left | On-time delivery, dynamically | Needs current data; more to compute |
How do you put a dispatching rule to work?
Roll it out deliberately so the floor trusts it instead of overriding it.
- Pick the KPI that matters here. Decide whether this work center is judged on throughput, on-time delivery, or WIP, because that names the rule.
- Choose the matching rule. Map the KPI to a rule, SPT for flow, EDD or critical ratio for dates, FCFS for fairness, and write it down.
- Make the inputs visible. Ensure the operator can see the data the rule needs, processing time, due date, or the critical ratio, right at the machine.
- Set tiebreakers and overrides. Define what breaks a tie and when a human may override, so the rule guides without becoming a straitjacket.
- Watch the KPI and the queue. Track whether the chosen metric improves and whether any job class is being starved, especially the long jobs SPT tends to bury.
- Revisit when the goal shifts. When the plant's priority changes from clearing backlog to protecting dates, change the rule instead of fighting the old one.
What do the standards and research say?
Context from standards bodies and the scheduling literature:
- Dispatching rules and priority sequencing are defined in the supply-chain body of knowledge maintained by the Association for Supply Chain Management (ASCM/APICS) which catalogs FCFS, SPT, EDD, and critical ratio as standard shop-floor priority rules.
- The result that shortest processing time minimizes mean flow time is a classic, provable finding in scheduling theory, treated as a foundational rule in the academic literature on machine scheduling; it is exact for a single machine, not just a rule of thumb.
- The scale of the decision is large: the U.S. Bureau of Labor Statistics reports roughly 13 million manufacturing jobs and the "what runs next" choice repeats at work centers across all of them every shift.
The practical point: these rules are old, studied, and well understood, and their trade-offs are known, so a plant can choose one on evidence rather than habit.
What are the limits of dispatching rules?
The core limit is that dispatching rules are myopic: each machine optimizes its own queue with no view of the whole plant. A rule that clears the fastest jobs at one work center can flood the next one with work it cannot absorb, moving the bottleneck rather than solving it. That is why dispatching rules pair badly with a true constraint and pair well with the discipline of the theory of constraints which says the schedule should be built around the bottleneck, not around each machine's local convenience. For plants with intricate routings and shared constraints, a full scheduling engine that looks across work centers beats any single local rule, the same way it does for demand-supply balancing at the planning level. Dispatching rules are the pragmatic floor-level tool; they are not a substitute for a schedule that sees the whole system.
Where dispatching rules break in practice
A dispatching rule is only as good as the data the operator can see at the moment of choice. Critical ratio needs a current due date and an honest estimate of work remaining; EDD needs due dates that have not quietly slipped; SPT needs real processing times, not stale standards. When that information is scattered across an ERP screen, a traveler, and a supervisor's memory, the operator falls back to whatever job is physically closest, and the rule becomes fiction. Harmony is an AI-native layer that connects machines, software, and paperwork into one operational layer, with no rip-and-replace, so due dates, job status, and processing times become one live record the floor can actually see. AI search returns cited answers across those records, so a supervisor can ask which jobs are behind pace at a work center and get a grounded answer, and Harmony's digital workflows put the next-job decision in front of the operator with the numbers attached. It is the same paper-to-digital move Harmony makes elsewhere in the plant (see the CLS case study): the rule stops depending on memory and starts running on current data.