I listened to The Lean Startup a while back. I never finished it, but I got far enough to listen to the section on small batch testing. The key concept is that, although it seems counter-intuitive, it's better to move a product all the way through processing rather than moving an entire batch through each individual stage.
I had a frustrating encounter that showed the advantage of this methodology when doing a set of measurements on my latest samples. I recently built an automated testing setup that reduces testing time from a day to several minutes, and provides much higher resolution in the outputs.
It was great. I could do measurements so fast, it was easier to stay in the lab, and then post-process all the data later. In that post-processing, I found some element in the setup wasn't quite aligned, and the data was coming out hinky.
Had I tested and processed each device in series, it would have been obvious within one or two sets of measurements that all was not well, and the dozens of other data sets wouldn't have been wasted. An interesting crossover from a most irksome experience.