Shop Floor Scheduling Software: Dispatching What Runs Next

Planning answers 'when will this job ship?' Dispatching answers 'what do I run next on this machine?' Big-system schedulers try to answer both with a finite-capacity optimizer, and in small shops that's the most-abandoned module in the category — reviews of major systems call it a feature that 'just plain doesn't work,' and one well-known legacy system locks scheduling to a single user.

Dispatching is simpler and more robust. Each work center gets a queue of ops in priority order; the operator works the top card; the lead reorders when reality changes. It survives bad run-time estimates because it never pretends to know the minute a job will finish.

Earliest due date is the default for good reason: sort each queue by the job's promise date and you'll rarely be surprised. Its weakness is jobs with many ops left — they look fine until it's too late. Critical ratio fixes that: time remaining until due divided by work remaining. Under 1.0 means the job is behind; sort ascending and the jobs in real trouble rise.

Two practical overrides belong on top of any rule. Group by setup when it saves real time — running three jobs that use the same fixture back to back can beat strict due-date order. And let a human pin a hot job, but show the cost: which other promise dates break and by how much. An expedite with a visible blast radius is a decision; an expedite without one is a surprise for three other customers.

Three views. The operator's: my machine's next three ops, with the traveler and setup notes one tap away. The lead's: every work center's queue side by side, drag to reorder, red for late and yellow for at-risk. And the wall: a TV in the shop showing the same board, updating itself — shops already jury-rig this with legacy software and a flatscreen.

Two lanes that small-shop schedules usually forget: outside processing (jobs at the plater or heat treater, with due-back dates and an overdue flag) and waiting-on-material. Those two lanes explain most 'where's my job' questions.

A schedule that someone re-types every morning will be abandoned. The board has to update from things the floor already does: scanning a QR code on the traveler to start an op and tapping complete with good and scrap counts. That's enough to move cards, color lateness, and keep promise dates honest.

Machine down? Mark the work center down and the affected ops should flag themselves for re-sequencing. Machine monitoring can refine this later, but it isn't a prerequisite.

If you're not buying software yet, upgrade the whiteboard: one column per work center, one magnetic card per op with job number, part, quantity, and promise date, cards sorted by due date top to bottom, a separate column for OSP with due-back dates, and a red magnet for anything late. Re-sort once a day at a fixed time.

Traveler is being built to replace exactly that board: a read-only late-and-next list first, then a drag-and-drop board with TV mode, an OSP lane, machine-down flags, and an expedite preview that lists the promise dates a move would break. It deliberately won't ship a finite-capacity optimizer. Planned pricing is $109/month flat with unlimited users.