Love it

...

the, the way the simulation works, just if people aren't familiar with Monte Carlo, um, it's basically you take something that's very complicated and impossible to simulate from first principles, uh, accurately, and you say, "Well, okay, well, we're gonna model it statistically." So, um, the way that works in practice is you've got, like, how long is it gonna take to make this part? Um, well, a person can say, "You know, I, I'm not sure, but I think it's gonna take, you know, five to ten minutes, um, to weld that together." It's like, okay, well, we could either, either produce a distribution on that style of part analytically and build a statistical model, or we can say, "All right, five to ten minutes, so let's put a uniform distribution on that." And then what happens is you, you build a model, um, that reflects every operation in production, uh, and then you simulate it end to end and say, "Okay, well this time it took five minutes.

This time it took ten minutes." You simulate 1,000 times, and then at the end of the day you have an outcome that says, "Hey, this part shipped on time 85% of the time." Um, what can we do? How can we change the scheduling heuristics to nudge that number up, you know, to 90% or 95 or, or 100%? Um, there's always, like, late scrap and stuff, and you model that too.

Um, so there's, uh, there's the statist- statistical part, um, and then there's, um, what I call, you know, part selection heuristics, and that is what do I work on next? And that question is asked 1,000 times at every manufacturing operation every day.

So that's an impossible thing to, to answer optimally-

Love it

...