Production bottleneck: how to find it, and what to do about it

A bottleneck is the step in your process with the least capacity relative to the work sent to it. Everything downstream waits for it, and everything upstream piles up in front of it. The plant as a whole can never produce more than the bottleneck can process, however fast the other machines run.

That has an uncomfortable consequence. An hour gained on a machine that is not the bottleneck produces nothing extra: the parts just wait longer in front of the bottleneck. An hour lost on the bottleneck is an hour lost for the entire plant, and it never comes back.

Picture three steps in a row. Cutting can do 60 parts an hour, welding 25, painting 40. The line produces 25 parts an hour, full stop. Speeding up cutting to 80 changes nothing. A welding station standing idle for one eight hour shift costs the plant 200 parts, and no other step can make them up.

Most plant managers can name a machine they think is the bottleneck. Often they are right. Sometimes the bottleneck has moved, because the product mix changed, and the plant is still protecting the old one. Four signals help you check.

For each work center, add up the hours of scheduled work for the coming week, including setups, and divide by the hours it has available on its shifts. The work center with the highest ratio is the bottleneck for that week. This is the most reliable method, because it is based on the orders you actually have. If you need help with the available hours, read how to calculate machine capacity.

Walk the floor and look for work in progress. The biggest pile of parts waiting is usually in front of the bottleneck. The machine right after it is often idle or starved.

Take the orders that shipped late last month and trace where they spent the most time waiting. If most of them waited in front of the same work center, that is your answer.

When the bottleneck goes down, the whole plant feels it within hours. When a non bottleneck goes down, the plant often catches up without anyone noticing.

The bottleneck is a property of this week's orders, not of the machine. A plant making mostly housings may be limited by the mill that cuts them. Next month, with more shafts in the mix, the lathe may be the constraint. That is why a finite capacity schedule, which recalculates load from the real order book, finds the bottleneck better than a rule of thumb.

In the machine shop example on our home page, twelve orders run through six work centers in one week. CNC Mill 2 carries 82 percent of its available hours, while no other work center goes above 51 percent. Both late orders are housings, and both wait for CNC Mill 2. The schedule does not just say the mill is busy. It says which orders it will make late and by how many days, which is exactly what a planner needs to decide what to do.

The order matters. Start with the cheap options that recover lost time on the bottleneck, and buy capacity only when they are used up.

Make sure the bottleneck always has the next job ready, with material, tooling and drawings at the machine before the current job finishes. Stagger breaks so it runs through them. Schedule it first and plan the other work centers around it, not the other way round.

Every minute of setup on the bottleneck is a minute of plant output lost. Group similar jobs together on the bottleneck, as long as that does not push another order past its due date, and move setup preparation offline so the machine only stops for the change itself.

Inspect parts before they reach the bottleneck, not after. Scrapping a part after the bottleneck worked on it wastes the scarcest time in the plant.

Look for operations that another machine can do, even if more slowly. A slower machine with spare hours can take jobs away from the bottleneck and raise total output. In a scheduler, this means setting alternative work centers on the operations that allow it.

Only after the first four: a second or third shift on the bottleneck alone, an extra operator, overtime, outsourcing some operations, or finally a new machine. Because you know it is the bottleneck, you know the investment raises plant output rather than moving a queue somewhere else.

Relieve one bottleneck and another work center becomes the constraint. That is normal and it is progress: the plant now produces more, limited by a different step. Keep measuring load against capacity every week, because the constraint will keep moving with your order mix.