Capacity expansion is the expensive answer to a throughput problem. Before buying another machine or adding a shift, it’s worth knowing how much of your existing capacity is being lost to sequencing decisions — because in most plants the answer is a surprising amount.

Why scheduling is genuinely hard

This isn’t a criticism of your planners. Production scheduling is a mathematically difficult problem. With twenty jobs across six work centres, the number of possible sequences exceeds anything a person can evaluate. Add changeover times that depend on which job preceded which, machines that can run some products but not others, operators with different skill sets, material arriving on different dates, and customer priorities that shift mid-week, and you have a problem no spreadsheet solves.
So planners do what’s rational: they find a workable sequence, not an optimal one. Workable is usually good enough to ship. It’s just rarely efficient, and the gap between workable and optimal is where your unused capacity is sitting.

Where the Recovered Hours Come From

Changeover sequencing is the largest single source in most plants. If a colour change from light to dark takes twenty minutes and dark to light takes ninety, sequencing matters enormously. Across a week of jobs, grouping intelligently can free hours that were previously invisible because nobody counted setup as lost production.

Bottleneck protection is the second. Every plant has a constraint. If that machine is idle for any reason — waiting for material, waiting for the previous stage, waiting for an operator — the whole plant lost that time and can never recover it. A scheduler that treats the constraint as sacred and sequences everything else to keep it fed produces meaningfully more output from identical equipment.

Batch sizing is the third. Larger batches mean fewer changeovers but longer lead times and more work in progress. The right answer varies by product and by how loaded you currently are, which is exactly the kind of conditional judgement that gets frozen into a fixed rule and then never revisited.

Reacting to Reality

A schedule’s real test is what happens when it breaks, which is daily. A machine goes down, a rush order arrives, a material delivery slips. Manual rescheduling under time pressure means patching the immediate gap and accepting whatever knock-on effects follow. Automated rescheduling can re-optimise the remaining sequence in seconds and show you the consequence of each option before you commit.

Making it Stick

The failure mode here is well documented: a system produces a mathematically optimal schedule that ignores something real — an operator’s certification, a tool that’s away for repair, a customer who must be prioritised for reasons outside the data. The schedule gets overridden once, then routinely, then ignored.
Avoid this by capturing constraints properly before go-live and by giving supervisors a straightforward way to lock or override specific jobs with a reason. Those reasons are also the best available source of constraints you didn’t know you had.
ticktick.ai schedules against real constraints — changeover matrices, machine capability, material availability, labour — and reschedules when conditions change.