We use a system dynamics simulation model based on the Factory Physics perspective to study the cumulative effect of continuous improvement in arrival variability, process variability, defect rate, time to failure, repair time and set-up time on operating curves in a flow-shop environment. We find that small, simultaneous improvements at multiple locations in the line can provide reductions in cycle time comparable to, or sometimes superior to, those obtained by a major improvement at a single location. The reduction of process variability is often an excellent alternative to reducing the mean of a parameter, because process variability at a given station affects not only that station, but all downstream stations due to its impact on flow variability. Improvements at non-bottleneck stations also have significant benefits, again by reducing the variability of flow to the bottleneck station. These results suggest a broader interpretation of some aspects of the Theory of Constraints, and help explain the successful results of the Toyota Production System.