[Inventory is measured by] the money the system invests in purchasing things the system intends to sell. - Eliyahu M. Goldratt

Well, since we started keeping data on the bottlenecks, I’ve been noticing I’m able to predict several weeks in advance what each bottleneck will be working on at a particular time. See, as long as I know exactly what’s in queue, I just take the average setup and process times for each type of part, and I’m able to calculate when each batch should clear the bottleneck. Because we’re only dealing with one work center, with much less dependency, we can average the statistical fluctuations and get a better degree of accuracy.

My system does not make sense at all, but by God it’s working.

When a proto-type — a new initiative — doesn't work, we face two alternatives: one is to bitch about reality and the other is to harvest the gift it just gave us, the knowledge of what has to be corrected.