Production scheduling in fresh food manufacturing – an anonymised case
Case summary
A fresh food manufacturer already planned demand with Horizon. This case is about what comes after the forecast: production scheduling.
Scheduling used to be manual. Planners sequenced production orders from experience, trying to keep cleaning, changeovers and setup times down while meeting delivery dates. As volumes and product variety grew, that became harder to sustain: schedules took longer to build, were hard to check, and often changed at the last minute on the shop floor.
The aim was an executable, data-driven schedule, without losing what the planners and operators know. Today, 90% of scheduling is automated. The planners work on the part that needs them.
Company context
Fresh food means short shelf life, strict hygiene and frequent changeovers between products. Production runs on lines with their own capabilities and limits. Sequence decisions directly affect cleaning time, setup losses and how stable the day runs. Small mistakes in assumptions show up on the shop floor quickly.
From assumptions to explicit scheduling data
The first phase was about data, not optimisation. Cleaning and changeover logic had to be made explicit: instead of general rules, detailed matrices of how long it takes to switch from one product or recipe to another. Setup times were refined too; several assumed durations did not match reality, and that only became visible once schedules were shown line by line.
Data went in, schedules came out, issues were discussed and assumptions corrected. Each round made the schedule more realistic. The customer’s feedback also shaped how the schedules are generated and shown.
How the schedule is built
Horizon builds minute-by-minute sequences per line with finite capacity scheduling and mixed-integer programming, on the same data as the demand plan. Cleaning and changeover rules are applied the same way every time, whoever builds the schedule. More on production scheduling in Horizon.
Using it in the factory
Reality changes during the day. Planners needed to adjust quickly without rebuilding the whole schedule, so the last phase focused on use: how planners change a schedule, how changes show up, and how it stays executable on the shop floor.
Where they are today
Manual sequencing, guesswork and rework are down. Cleaning and changeover logic is explicit rather than in people’s heads, and schedules are consistent across planners. People still oversee the schedule; they no longer build it by hand.
