Constraint Thinking: Why Keeping Every Station Busy Can Hurt the Whole System
One slow step caps how much a line can make. Keeping every other step busy often just piles up unfinished work and makes orders take longer.
One step sets the pace
Picture a line where every part goes through four steps. Three steps can each handle 15 parts an hour. One can handle only 10. The line can never ship more than 10 an hour, no matter how hard the other three work. That slow step is the constraint. Keeping every station busy doesn’t, by itself, make the line any better. In general, when each step feeds the next, the total output (throughput) is capped by a constraint. It might be a machine or a team. It might also be a rule, a shortage of material, missing information, or a mix of these.
Running every other step flat out often just builds up extra work-in-progress (WIP): unfinished work piled up in front of the slow step. You can end up with more stuff sitting around, orders that take longer, and a harder time seeing what’s going on. And none of it ships a single extra part.
Keep work flowing, not just busy
A simple rule called Little’s Law ties these together: WIP = Throughput × Cycle Time. Cycle time is how long one piece of work spends in the line, start to finish. It works like basic bookkeeping for flow. Think of a lunch line where the servers don’t speed up: the longer the line, the longer each person waits. In the same way, if WIP goes up and throughput doesn’t, the average cycle time has to go up. So the math says it’s costly to start extra work just to keep everyone busy.
Here’s a better question to ask. What stops the slow step from turning out useful work? And which ups and downs, like breakdowns or late parts, should we cushion so the slow step keeps working?
From idea to software
A good decision tool should put a few things in plain sight. How much work is coming in? How much can each step handle? How much do things bounce around? And where is the slow step? It shouldn’t bury those answers in a dashboard. Interactive models help because you can try a change and see what it does. Does it move the constraint for the whole system? Or does it just make one local number look better?