Dispatching: Three Jobs Waiting, One Machine — Which Rule Wins?

Monday, 07:00. Three jobs wait at the lathe, and job P is already a day overdue. The foreman runs it first. It feels right: it's the job sales keeps calling about. The numbers disagree.

• P: 3 days of work, due yesterday (day −1).
• Q: 2 days of work, due day 2.
• R: 1 day of work, due day 3.

Most overdue first (P, Q, R): P finishes day 3 (4 days late), Q day 5 (3 days late), R day 6 (3 days late). Three late orders, 10 days of lateness in total.

Q, R, then P: Q finishes day 2 (on time), R day 3 (on time), P day 6 (7 days late). One late order, 7 days in total.

The foreman turned one unhappy customer into three. Whether P's customer is worth that is a real business decision. It should be a decision, not a reflex.

Manufacturing dispatching is choosing which job a machine runs next from the jobs waiting in front of it, using a clear rule. The schedule says what should happen; dispatching decides what happens at the machine when a job finishes and reality doesn't match the plan. How that schedule is built is covered in production scheduling.

The four dispatch rules everyone should know

• FIFO (first in, first out): run jobs in the order they arrived. Simple, blind to due dates.
• EDD (earliest due date): run the job due soonest.
• SPT (shortest processing time): run the quickest job first.
• Critical ratio (CR): time left until the due date ÷ work left. Below 1 means the job is behind; above 1 means it has slack.

FIFO vs EDD vs SPT: one machine, four jobs

Four jobs are waiting at hour 0, listed in arrival order. Lateness counts hours past due; early jobs count as zero.

FIFO (A, B, C, D): A ends at 6, B 8, C 16, D 19. Late: B by 2, D by 4. Average lateness 6 ÷ 4 = 1.5 h, two late jobs.

EDD (B, A, D, C): B ends at 2, A 8, D 11, C 19. Late: C by 1. Average lateness 1 ÷ 4 = 0.25 h, one late job.

SPT (B, D, A, C): B ends at 2, D 5, A 11, C 19. Late: A by 3, C by 1. Average lateness 4 ÷ 4 = 1.0 h, two late jobs.

EDD wins on lateness. But SPT has a trick: its average completion time is (2 + 5 + 11 + 19) ÷ 4 = 9.25 h, against 10 h for EDD and 12.25 h for FIFO. SPT gets jobs out the door fastest, which shrinks the queue at the next machine. Its dark side: a long job can be pushed back forever if short ones keep arriving.

FIFO came last on both counts.

Notice that the foreman's "most overdue first" is EDD. EDD works well while jobs can still make their dates. Once a job will be late no matter what, EDD lets it drag others down with it, as P did.

Critical ratio: triage for jobs

An emergency room doesn't treat patients in arrival order, and it doesn't treat the loudest one first. It treats whoever will get worse fastest without care. Critical ratio does the same for jobs.

Three jobs at a welding cell, today is day 0:

• X: due in 4 days, 5 days of work left, including later operations. CR = 4 ÷ 5 = 0.8
• Y: due in 3 days, 3 days of work left. CR = 3 ÷ 3 = 1.0
• Z: due in 10 days, 4 days of work left. CR = 10 ÷ 4 = 2.5

X runs first: it's already behind. Y is exactly on track, with no room to wait. Z can wait days without harm, even if it arrived first and FIFO would have picked it.

Critical ratio is the best single rule when jobs have long routes, because "work left" counts the operations still ahead. Its weak spot: it needs more data, and it's only as good as your routing times.

Myth: there is one best dispatch rule

There isn't. EDD minimizes lateness, SPT minimizes average waiting, CR balances both for multi-step routes, and grouping similar jobs can beat all three on a machine with long changeovers. Choose a rule per work center based on what hurts most there, and write it down. A written rule followed every shift beats a clever one used only when the right person is on duty.

Who may override the sequence, and how to log it

Rules need exceptions. A customer escalates, a die breaks, material for the top job is missing. What matters is that overrides are deliberate:

1. Name who can override. Usually the shift supervisor within the shift and the planner beyond it. Operators shouldn't reshuffle the list to dodge an awkward setup.
2. Log every override in one line: job moved, job displaced, reason, who decided. "Moved P ahead of Q, customer escalation, approved by J. Silva."
3. Review the log weekly. If the same reason shows up ten times, the rule or the data is wrong. Fix that instead of overriding forever.

The log also explains your OTIF numbers. When a customer asks why their order slipped, "job P was prioritized after an escalation" is a much better answer than a shrug.

How it looks in factory.online

In factory.online, the job orders for each machine come from the finite schedule in sequence, so the operator isn't guessing from a stack of papers; see what actually goes to the shop floor. Output confirmations record what really ran and when, so the planner can compare the planned sequence with the actual one and spot where overrides happened.

If your shop floor argues about "what's next" every morning, book a demo and we'll walk through it with your own queue.