How to find your factory bottleneck without measuring everything

There's an uncomfortable truth in every plant: your capacity is not the sum of your machines. It's the capacity of your most loaded resource. Improving anything else - buying a faster machine, adding overtime on a shift that isn't it - doesn't add a single unit to your daily output. It only grows the pile in front of the bottleneck.

That's why the first question in any capacity improvement isn't "what do I buy" but "where is my bottleneck". And the good news: in an SME, the bottleneck can be found in one morning.

The 3 signals that give the bottleneck away

Before measuring anything, walk the floor with these three questions:

• Where does material pile up? Inventory waiting in front of a resource is the most reliable signal. A queue can point to a constraint, but batch policy, release timing and outages can also create one.
• Which resource never stops? A resource with high sustained utilisation is a candidate; confirm with time and demand data. If a machine "can't even take a bathroom break", you have a candidate.
• Which resource, when it stops, stops the whole plant? Ask production: "which breakdown ruins your whole day?". The answer provides a useful lead, not a measurement.

If all three signals point to the same place, investigate that resource first. If they point to different places, run the calculation below.

The 20-minute calculation

For each candidate resource you only need two numbers: the effective hours it has and the hours demand asks of it.

• Capacity: shifts x hours x days, multiplied by an honest OEE (if you do not measure OEE, record it as unknown and collect a short baseline rather than treating an assumed value as fact). A machine on two 8-hour shifts with 0.7 OEE gives you about 11 effective hours a day, not 16.
• Load: annual units of each part number x its cycle time on that resource, added up. Demand converted into hours.
• Utilisation: load / capacity. Interpret utilisation alongside variation, planned downtime and demand. There is no universal safe threshold; the most constrained resource needs a closer look.

The resource with the highest utilisation rules. The whole factory should organise around protecting its hours: never short of material, its breakdowns fixed first, its changeovers done with method.

And once you find it, then what?

The levers, in order of cost:

• Free: protect the bottleneck. No stops for lack of material or schedule. Its time is the whole factory's time.
• Almost free: SMED on the bottleneck (recovered time only creates sales when there is demand and the rest of the flow can support it) and offloading work another resource can do.
• Cheap: overtime ONLY on the bottleneck. On any other resource, it's money thrown away.
• Expensive (and last): more shifts or another machine. Only when the three above are squeezed dry - if you buy capacity on the wrong resource, the bottleneck moves and you've paid for nothing.

And watch the last effect: when you truly improve the bottleneck, it moves to another resource. That's not failure, it's the signal that it's working. Walk the floor again and start over.

The question that does need numbers

"Can we take this new order?" is the serious version of this exercise: it's not enough to know where the bottleneck sits today - you need to know what happens if demand rises 15% or you add 8 hours of overtime. That isn't solved by walking: it's solved by crossing demand and capacity resource by resource and simulating before you say yes to the customer.