It's Not Luck by Eliyahu M. Goldratt Read Free Book Online Page A

days.”
    “I don’t get it at all,” Stacey admits. “But first, how do these changes lead to better fulfillment of orders with less stock. I don’t see the connection.”
    “It’s simple,” I interrupt. “It’s all a matter of statistics. Our knowledge of what a shop sells of each item is very rough. One day they can sell ten units of something and the next day zero. Our forecast is based on averages.”
    “That’s clear,” Stacey says.
    “Now, which forecast will be more accurate?” I ask. “The forecast of the sales of one shop or the aggregated forecast of the sales of one hundred shops?”
    “The aggregated forecast,” she answers.
    “You are right, of course. The larger the number, the more accurate the aggregated forecast. The mathematical rule is that as we aggregate more and more shops the accuracy of the forecast improves in proportion to the square root of the number of shops that we aggregate. You see, when Bob moved the majority of his inventory from twenty-five regions into the plants themselves, his forecast became more accurate by a factor of five.”
    “Alex, you with your statistics,” Bob cuts in. “I never understood them. Let me explain it in my way. Stacey, when you ship to a regional warehouse and you have, on average, three months’ inventory in the system, this inventory will be sold, on average, three months after the plant shipped it, right?”
    “Provided that you have produced the right stuff in the first place, otherwise it will be even worse,” she agrees. “Now I see; as long as the plants have shipped immediately whatever they produced, their shipments to the warehouses were based on the forecast of what will be sold in that region three months down the road. Knowing the accuracy of such forecasts, especially when you are dealing with over six hundred products, I can imagine what was going on.”
    “Don’t forget,” Bob adds, “that on top of six hundred and fifty products, I have twenty-five regional warehouses. This considerably adds to the mismatch.” We all nod, and Bob summarizes, “When a regional warehouse goes to fulfill a shop order, some items are always missing. At the same time, we do have these items; we have a lot of them, but in other warehouses. Now the madhouse starts. The warehouse manager is pressing the plants for immediate delivery, and if he can’t get it, he starts to call other warehouses. You won’t believe the amount of cross shipments between warehouses. It’s horrendous.”
    “I can easily believe it,” Stacey says. “What else can you expect when the plants ship the goods three months in advance of consumption? You must end up with too much of one product in one place and too little in another. So I see what you’ve done; you wiped out local considerations and decided to hold the stock at the origin—at the plant.”
    “Where the aggregation is the biggest,” I add. “Where the forecast is the most accurate.”
    “But you still need the regional warehouses,” Stacey says thoughtfully.
    “Yes,” Bob agrees, “since we want to respond quickly to the shops’ orders and cut shipping costs. Otherwise I’d have to ship each order to each shop directly from the plant. Federal Express would love it!”
    “I see,” she says. “So how did you determine how much inventory you need to hold in each regional warehouse?”
    “Ah-ha. That was the sixty-four-thousand-dollar question,” Bob beams. “Actually, it is quite simple. I just had to extrapolate from what we learned about buffering a physical constraint. Stacey, you’re probably as paranoid as I am about building inventory buffers before a bottleneck.”
    “Yes, of course,” Stacey agrees.
    “How do you decide on the size of a bottleneck’s buffer?”
    “We figured that out together already in the Bearington plant,” she smiles. “The size of the buffer is determined by two factors: the expected consumption from it, and the expected replenishment time to