Developed by Eliyahu M. Goldratt (The Goal, with Jeff Cox, 1984).
Raises a system's output by finding its one binding bottleneck, keeping it busy, pacing everything else to it, expanding it, and then finding the next.
A fulfillment line ships only as fast as its slowest station
Each box is a station with its capacity in orders an hour, and every order passes through all five. Raise any station except the bottleneck and output stays put; raise the bottleneck and output climbs until another station takes over. Illustrative.
Orders an hour
Orders shipped an hour40, set by PickPackCheckLabelLoad
- Pick. Pickers can pull 80 orders an hour. Adding pickers here only builds a longer line of carts in front of the packing benches.
- Pack. Two benches pack 40 orders an hour, the least of any station, so the whole warehouse ships 40. An hour lost here is an hour lost to every customer.
- Check. One scanning station checks 60 an hour. It waits on packing today, and it is the next station in line to become the bottleneck.
- Label. Labeling handles 70 an hour and stands idle part of every shift. That idle time is normal at a station that is not the constraint.
- Load. The dock loads 90 an hour. Speeding it up would change nothing until every station before it can keep up.
An example
A distributor of replacement parts for restaurant kitchens shipped about 40 orders an hour at its busiest, and the list of late orders grew every afternoon. The operations manager had already added two pickers, and output had not moved. She walked the floor at three o'clock and found carts of picked orders lined up in front of the two packing benches, while the scanning station past them stood idle a third of the time.
Packing was the constraint. The team first made every bench hour count: a picker folded boxes and stocked fill during slack time, and the packers stopped taking their breaks together. Pick lists were then released at the rate packing could take them, which cleared the carts off the floor. A third bench came last. Output rose to 60 orders an hour and stopped there, because the single scanning station had become the new constraint. The next quarter's budget went to a second scanner.
Why it works
Every order passes through every station, so the line can ship no more than its slowest station handles. An hour gained anywhere else adds work in progress and nothing more; an hour lost at the constraint is lost to the whole line, and no other station can make it up. That one fact turns a vague push to improve everything into an order of work: protect the constraint, pace the rest to it, and only then pay to expand it.
The order matters. Exploiting the constraint before buying capacity is cheap and often frees a large share of it, since constraints lose time to setups, breaks, and defective work from upstream. The last step keeps the method honest: once the constraint is expanded, another stage limits the line, and the effort has to move with it.
The thinking behind it
Eliyahu M. Goldratt Eliyahu M. Goldratt, an Israeli physicist who turned to management, set out the theory of constraints in The Goal (1984), a business novel written with Jeff Cox. Its argument is that a system's output is limited by its constraint, so improvement means finding the constraint, exploiting it, and elevating it, with everything else paced to it, and then looking for the next one. He kept teaching through novels: It's Not Luck (1994) added his thinking processes, and Critical Chain (1997) carried the theory into project scheduling.
How to run it
- Map the flow. Draw every stage the work passes through, from order to shipment, with each stage's capacity in the same unit, such as orders an hour.
- Find the constraint. Look where work piles up, where people expedite, and where overtime gathers; the stage just past the pile is usually the one.
- Exploit it. Remove every loss at the constraint first: idle breaks, long setups, waits for materials, and defective work arriving from upstream.
- Pace everything else to it. Release work into the line at the constraint's rate, and let other stations sit idle before they build piles in front of it.
- Elevate it. Only then spend on more capacity at the constraint: another bench, a second shift, or work sent outside.
- Find the next one. When output stops rising the constraint has moved, so go back to the second step and retire rules written for the old one.
Where it goes wrong
The usual failure is improving the station that is easiest to improve. Capacity added upstream of the constraint buys a bigger pile and the same output; spot it when work in progress rises after an investment and shipments do not. Measure output at the end of the line before and after every change.
The second is chasing utilization. A plant that keeps every machine busy floods the constraint and hides it under inventory. Idle time at a station that is not the constraint is normal, and a target that punishes it should go.
The third is stopping after the first win. Rules built around the old constraint, such as a release schedule or a staffing plan, outlive it and become the new limit. Recheck where work piles up after every change.
When to use something else
- Value Stream Mapping. When the flow itself is unclear and the waste sits in waits and handoffs across the whole line, before any one constraint can be named.
- Theory of Constraints Thinking Processes. When the constraint is a policy or a measure, such as an approval rule, and no added capacity will lift it.
- Queueing Theory. When arrivals are uneven and the question is how long work waits in line at a busy stage.
Where it sits in SOLVE eX
Go to the source
- The Goal, with Jeff Cox, North River Press, 1984