Make-to-Order Shops: Scheduling Backward From the Due Date

Make to order backward scheduling pulls each job to finish just in time, so parts are made close to when they ship rather than as early as a machine happens to be free. A make-to-order shop's commitment is the due date, and building weeks ahead only clogs the floor with work in process. EDGEBIC by User Solutions places a backward job latest-feasible against the same finite capacity as forward scheduling, and when a date genuinely will not fit, it falls back to forward and tells you before you promise.

For the mechanism on its own, see backward scheduling explained and forward versus backward scheduling. For the sector view, see make-to-order scheduling, and pair this with contract manufacturer scheduling and high-mix low-volume scheduling. The full map is at how different industries use EDGEBIC.

The early-start problem in a make-to-order shop

A make-to-order shop takes orders it does not stock, each with a promised date, and builds them from raw material through a routing. The natural instinct of most schedules is to start everything as early as possible: the moment a machine is free, load the next job. It feels efficient. It is not.

Starting early means a job due in three weeks gets cut today because a machine was open, while this week's due orders wait behind it. The floor fills with half-built work that did not need to run yet. Cash sits in work in process. And when a rush order lands, there is no room, because the machines are busy on jobs that were not due for a fortnight.

The commitment in a make-to-order shop is the due date, not the start. A schedule that respects that builds each job to land on its date, not to start as soon as possible.

Backward scheduling: finish just in time

Backward scheduling inverts the default. Instead of pushing work forward from today, the engine places the whole order so its last operation ends at or before the due date, and every upstream step is right-aligned to feed its successor. The slack ends up in front of the job, not after it.

The effect on a make-to-order floor is direct. A job due in three weeks is scheduled to start late enough to finish on time, so it does not consume machines this week. This week's due orders get the capacity they need. Work in process drops because jobs are cut and machined close to when they ship. The floor clears, and when a rush arrives there is room to place it.

You turn backward on per order, and you can make it the site default so every new order inherits it. The engine places steps in reverse order, each deadlined by when its successor actually got scheduled, using the same allocator as forward scheduling. Shifts, holidays, downtime and machine capacity are enforced identically. A machine still runs one job at a time. Backward only changes where the engine searches for a slot: right-aligned to a deadline instead of left-aligned to an earliest start.

A worked make-to-order job

Consider a 40-piece order on a three-step routing, due at the end of a Friday, scheduled backward. Machining takes 22 hours, deburr 8, inspection 4, on single-shift day work.

Scheduled backward, the engine places inspection last, ending in time to make the Friday deadline, then deburr ending in time to feed inspection, then machining ending in time to feed deburr. The whole job lands so the last operation finishes by Friday's close, and the start falls where it needs to, several days before the due date rather than the previous week. Three days of slack sit in front of the job instead of after it, and the shop is not holding a half-built order for a week first.

Compare that to the forward default, which would start machining as soon as a machine opened, finish the whole job early, and let it sit as finished or in-process inventory until the Friday it ships. Same routing, same capacity, very different floor.

When a date will not fit: the honest fallback

Backward scheduling makes a promise: finish by the due date. Sometimes the capacity is not there, and the shop needs to know that before it commits, not after.

When a backward job cannot fit, whether the date is too close, a step would start before today, or capacity forces an overshoot, the engine schedules the whole order forward instead and records the miss. It places the order as early as possible from today, computes the realistic finish, and flags the job as late with the number of days.

On a schedule run you can be prompted with that result: the forward window showing where the job would start and finish, whether it is late and by how many days, and editable due date, start date and priority. Nothing commits until you accept it. You can accept the forward fallback as the best that fits, adjust the date or priority and re-run, or cancel and leave the plan untouched. Either way, you learned the date was unachievable before you told the customer, which is exactly when that information is worth something.

The trade backward makes, and when to keep forward

Backward scheduling buys lower work in process by removing downstream slack. That is a real trade. A job pulled tight to its due date has no runway: if a machine goes down, the job is immediately at risk, because there is no buffer after it.

So backward is opt-in per job, and that is deliberate. Use it where cutting work in process matters and the routing is reliable. For jobs where you want the protection of finishing early, keep them forward, which front-loads the work and leaves slack before the date so a disruption still has room to recover. A mixed shop runs both: backward for the steady, well-understood work, forward for the jobs where early completion is worth the inventory. And even a backward job that cannot fit rolls back to forward automatically, so the downside is bounded.

For shops where a bottleneck resource drives the plan, backward scheduling coexists with bottleneck anchoring under a clear precedence, covered in combining the two.

There is one more reason a make-to-order shop values the backward option: quoting. When a customer asks whether you can hit a date, a backward simulation places the job to finish just in time against your current load and tells you whether it fits before you commit. If it does, you have a start date and a defensible promise. If it does not, the same forward fallback shows you the realistic finish and how many days short you are, so you can negotiate the date, offer a priority bump, or decline rather than promising something the floor cannot deliver. That is the difference between a quote based on a wish and a quote based on the plan.

Common mistakes with backward scheduling

Turning it on with no real due date. The backward pass needs a genuine due date to right-align against. A job with an empty or placeholder date schedules forward instead. Give make-to-order jobs real dates.

Expecting a started job to re-pull backward. Backward decides where a new job is born. Once a job has been scheduled and started, a reschedule moves it forward from its resume point rather than re-pulling it, because the work already began. Plan the backward placement when the order is created.

Using backward on fragile routings. A job with an unreliable operation pulled tight to its date has no recovery room. Keep those forward for the slack, or address the reliability first.

Ignoring the fallback flag. When a job falls back to forward and is flagged late, that is the signal to act, not to accept quietly. Read the forward window and decide what gives before you promise.

Rolling it out

1. Decide where backward fits: steady, reliable jobs where cutting work in process matters most.
2. Turn on backward scheduling for those jobs, or make it the site default so new orders inherit it.
3. Confirm every backward job has a real due date, so the engine can right-align against it.
4. Run the schedule and read a backward job end to end, checking that it finishes on its date and starts later than the forward default would.
5. Set the fallback to prompt you so a job that cannot fit surfaces the forward window and editable dates before it commits.
6. Keep fragile or protection-critical jobs forward, so they retain slack before the due date.

Bring a batch of make-to-order jobs and their due dates to a demo of job shop scheduling software, and we will run them backward and show you where the floor clears.