In almost every small manufacturing plant there is one piece of equipment that decides how much you can make. Find it, schedule it hour by hour, and let everything upstream and downstream serve it. An hour gained anywhere else on the floor is worth nothing; an hour gained on that machine is an hour of extra output for the whole plant.
This is old thinking and it stays true because it is arithmetic rather than a management style. What changes it from a slogan into a schedule is being specific about which machine, and then being disciplined about not optimising the others.
Find it honestly
The bottleneck is not always the machine people complain about. Three signals, in order of reliability:
- Work piles up in front of it and never behind it. This is the definitive test. Walk the floor at the end of a shift and look at where material is waiting.
- It is the only thing that never gets idle time. If every other station has gaps and this one runs continuously, that is your ceiling.
- Overtime is always scheduled around it. Plants discover their bottleneck through the overtime roster long before anyone names it.
In a soap plant it is usually the mixer or the extruder. In a food plant it is frequently the oven or the filler. In chemicals it is often not a machine at all but a curing or settling step — time in a tank that no amount of labour shortens. A time-based bottleneck is still a bottleneck, and it is the one most often left out of the schedule because nobody is standing at it.
Schedule the bottleneck, not the plant
The instinct is to build a schedule for every station. Do the opposite. Build a detailed schedule for the constraint and let the rest follow.
For the constraint, plan:
- Sequence, ordered to minimise changeover — light colours before dark, low-fragrance before high, the same base grouped together.
- Campaign length, long enough that the changeover cost is a small share of the run, short enough that you are not making three months of a slow-moving variant.
- Planned maintenance, on the schedule as a job, not squeezed into the gaps. The gaps do not exist on a constraint.
For everything upstream, the only rule is: keep a buffer in front of the constraint so it never waits. For everything downstream: have the capacity to take what the constraint produces without backing up into it.
Changeovers are capacity
On a constraint, a changeover is not a nuisance — it is lost production, and it comes straight off your output ceiling. Four hours of changeover a week on a machine running 50 hours is eight percent of your plant's total capacity.
Which means two things are worth real effort:
- Sequencing to avoid changeovers costs nothing but planning attention and often recovers several percent of capacity.
- Shortening the changeovers you cannot avoid — pre-staging materials, cleaning tools ready, the next batch's paperwork printed — converts setup time into run time.
Both of these are worth more than speeding up a non-constraint station, which just produces work-in-progress faster for the queue in front of the mixer.
Where the schedule meets reality
A schedule built from capacity alone breaks on materials. The batch sequenced for Tuesday needs a raw material arriving Thursday, and nobody knew until Tuesday morning.
That is why a production schedule is only as good as its link to what is actually in stock and what is actually on order. In Aourix, the planning board and the production schedule read from the same inventory and procurement data the rest of the workspace uses, so a batch whose materials are short is visible when it is scheduled rather than when it is started. It does not make the material appear — nothing does — but it moves the discovery forward by a week, which is usually enough to resequence instead of stopping.
The awkward part
Running a bottleneck properly means accepting idle time elsewhere, and that is culturally hard. A supervisor whose station is idle looks unproductive. Efficiency measured per station rewards them for making stock nobody needs.
If you measure each station on utilisation, you will get high utilisation and a plant that is no faster, plus a warehouse of work-in-progress. The measure that matters on a non-constraint is whether it ever starved the constraint. Everything else is noise.
Start with one week
Pick next week. Write the constraint's schedule hour by hour — jobs, sequence, changeovers, maintenance. Do not schedule anything else. At the end of the week, record two numbers: hours the constraint ran, and hours it sat waiting, with the reason for each wait.
That list of reasons is your improvement plan, in priority order, and it will be shorter and more specific than anything a broader review produces.