Picked up The Goal again this week. Fourth time through. The book is a novel about a manufacturing plant manager named Alex Rogo whose plant is about to be shut down. He spends the book figuring out, with the help of a physicist named Jonah, that the only metric that matters is throughput, that throughput is constrained by the slowest step in the line, and that improvements anywhere except the constraint are wasted.
Goldratt published it in 1984. The lessons have not aged.
Every consultant should re-read this every eighteen months. Every founder should read it once. The bottleneck is the only thing that matters and it is never where you think it is.
Three things I notice on this read that I missed on prior reads.
The protagonist resists the diagnosis. Alex is told repeatedly what the constraint is. He keeps trying to optimize other parts of the plant because they are easier to see, easier to measure, easier to take credit for. The book spends a long time on this resistance because it is the actual obstacle. Not the constraint itself. The unwillingness to accept that everything else is a distraction.
The constraint moves. After Alex fixes the first bottleneck, a new bottleneck emerges somewhere else. He keeps expecting the work to be done. The work is never done. The constraint is a function of the system, not a flaw to be removed.
The metrics that matter are the ones nobody on the corporate dashboard cares about. Throughput, inventory, operating expense. Three numbers. The plant is buried under fifty other metrics that look productive on slides and tell management nothing. The fifty are noise. The three are signal.
I have run engagements where the client could recite their dashboard from memory and could not tell me their throughput, inventory, or operating expense in a single number each. Every one of those engagements ended with us building the three numbers and watching the dashboard go quiet.
Goldratt's other contribution is the Theory of Constraints as a meta-method. Identify the constraint. Exploit it. Subordinate everything else to it. Elevate it. Repeat. It is the most useful five-step process I have ever found for systems work, and it generalizes far beyond manufacturing.