← All projects Independent build · Production scheduling

Production / IQ Scheduling Optimiser

An urgent order arrives. Can we take it without delaying any other customer? Test it in a scenario, see what is missing, compare the options and make the call.

CP-SATSchedulingCapable-to-promiseScenarios
+6 h 15 minlateness added by a 4-hour machine stop once re-planned, against +27 h 27 min pushed back
5 h 05 minless total lateness by running the urgent order first, with no extra late orders
Proven optimumboth breakdown recoveries, solved with the published plan's goals
4planning goals solved in a fixed order
1:16 walkthrough with narration and captions. Real software on a demo plant with synthetic orders, materials and machines.

The problem

A planned week rarely survives Monday. A customer asks for 60 housings by Tuesday afternoon, purchasing says the steel will be late, a machine stops. Each time the real question is the same: what happens to every other order if we say yes? In a spreadsheet that takes hours and still ends in a guess.

What it does

  1. Test an enquiryA new order goes into a scenario that never touches the order book, with a choice of how much the published plan may move.
  2. Check every materialEach missing material is listed with the time it is needed; the planner enters when each can really arrive.
  3. Wait or fill the gapIf an arrival is late, two schedules are solved: hold the machine for the order, or let other work use the wait.
  4. Split the deliveryFinds the largest quantity that still makes the date without any other order finishing later.
  5. Recover from a breakdownRe-plans around downtime and shows which orders move, by how much, against the published plan.
  6. Write the planA written plan per machine, with each idle gap explained and changeover times included.
An enquiry for 60 housings tested in a scenario: it can't be planned yet because 10 pcs of steel are short, and the tool offers to work out a part delivery.
An enquiry for 60 housings tested in a scenario: it can't be planned yet because 10 pcs of steel are short, and the tool offers to work out a part delivery.
Each missing material with the time it's needed. The planner enters when it can really arrive; here the steel is 10 h 40 min too late.
Each missing material with the time it's needed. The planner enters when it can really arrive; here the steel is 10 h 40 min too late.
The written plan for the shop floor: every idle gap is explained, here by material that is only ready to use at 14:00.
The written plan for the shop floor: every idle gap is explained, here by material that is only ready to use at 14:00.

How it decides

PipelineCritical orders first→Least weighted lateness→Fewest late orders→Least disruption

The optimiser (Google OR-Tools CP-SAT) solves the goals in that order. Each result is locked before the next goal is optimised, so a later goal can never trade away an earlier one. The planner can reorder or remove goals, weight or protect individual orders and cap lateness. An independent validator re-checks every schedule before it is shown.

Trade-offs it puts in front of the planner

SituationOption AOption B
Steel arrives 10 h 40 min lateOther work fills the wait: 26 h 15 min total lateness, 3 orders late, ships Wed 16:05This order first: 21 h 10 min, 3 orders late, ships Wed 10:00
M1 stops for 4 hoursPush everything back: +27 h 27 min total lateness, 8 orders lateRe-plan: +6 h 15 min, 3 orders late (same orders, machines, goals)

Running the order first cuts total lateness by 5 h 05 min with no extra late orders, and the customer gets Wednesday 10:00. Every figure here is read from the app, which shows each scenario's total lateness against the published plan. The software calculates the trade-off; the planner makes the call.

Late steel: both options solved side by side. Running this order first gives 21 h 10 min total lateness against 26 h 15 min, with the same three orders late.
Late steel: both options solved side by side. Running this order first gives 21 h 10 min total lateness against 26 h 15 min, with the same three orders late.
M1 stops for 4 hours: pushing everything back adds 27 h 27 min of lateness; re-planned by the optimiser it adds 6 h 15 min.
M1 stops for 4 hours: pushing everything back adds 27 h 27 min of lateness; re-planned by the optimiser it adds 6 h 15 min.

Built with

PythonFastAPIOR-Tools CP-SATReactTypeScriptVitepytestVitest

283 backend and 131 frontend automated tests. Scenarios are frozen snapshots of the plant, so nothing tested in them can be published by accident.

The video and all figures come from a demo plant with synthetic orders, materials and machines. No employer data is shown.

Planning should turn complexity into action.

Open to Production Planner, Materials Planner, Production Controller and Supply Chain Planning roles in the UK.

ContactStraight to my inbox

Send me a message

Recruiter, hiring manager or just saying hello: I’ll reply to the email address you give here.