PDF Summary:The Goal: A Process of Ongoing Improvement, by Eliyahu M. Goldratt

...

Take the advice of Ray Dalio in Principles: view yourself top-down as a machine, digesting inputs and creating outputs. From this perspective, you’ll be capable of studying and optimizing yourself.

...

*   In other words, this is money that is currently stuck in the system
*   Any investment that you can sell is inventory. 
    *   R&D that can be sold, like a patent, is inventory. In contrast, R&D to improve throughput can’t be sold and isn’t inventory.

• Operational Expense: the money the system spends in order to turn inventory into throughput
  • Includes labor, leases, R&D to improve throughput
  • It’s better not to consider where value is added - don’t get caught up over whether a dollar is investment or expense. If it’s not part of the packaged product being sold, it’s an operational expense.
• Ideally, you should try to improve all three at once.
  • For example, if you install robots, it should increase throughput by increasing sales, decrease inventory needed, and/or reduce labor costs.
• Be wary of a change that affects only one of these metrics - there may be second-order effects that backfire.
  • For example, reducing labor to reduce operational expense may decrease throughput.

If you make a subpart more efficient, you do not raise your competitive advantage if it does not increase output or reduce cost.

• Adding...

...

*   In a complex chain with many parts, this is difficult.

• In contrast, if X gets a boost. suddenly produces 10 units in 1 hour, Y can still only produce 5 units per hour, so the gain in X is stifled by the inability of Y to meet X’s boost.

As explained in the next sections, the solution is not to balance average capacity with demand, but rather to balance flow or throughput with demand. The way to do this is a drum-buffer-rope system, where the bottleneck determines the throughput and inventory of the entire system.

We’ll explain all of this with an analogy.

Analogy of the Hiking Line

Imagine a troop of 10 boys hiking single-file on a narrow trail in the woods. The leader of the pack sets a comfortable pace that everyone on average should be able to meet.

Every boy is only able to catch up to the boy in front - he can’t pass the boy in front. Thus, the speed of each boy is constrained by the boy in front.

Analogy to manufacturing: the first boy is the most upstream step; the last boy measures throughput; the distance in between is inventory.

Here’s an example of a negative fluctuation

• Say the 3rd boy in line stops to tie his shoes,...

...

• If the total throughput is a thousand dollars per hour, then the bottleneck is processing at a thousand dollars per hour, even if the literal operational costs or the parts going through it cost much less.
• Alternatively, take the entire operating expense of the factory, divided by the hours worked by the bottleneck - that’s the actual cost of the bottleneck.

In other words, a loss in the bottleneck means a loss to the entire operation, and should be viewed with such gravity

Other losses in effective throughput are also similarly costly. For example, feeding low-quality parts through the bottleneck will cause rejection later, leading to effectively lower throughput.

While time lost from the bottleneck can be made up for by hurrying non-bottlenecks, any extra effort here typically adds to operational expense (eg overtime pay). Ideally, the bottleneck is simply maintained at peak capacity at all times.

Identify the Bottleneck

The bottleneck is any resource whose capacity is equal to or less than the demand placed on it. Here are a few ways to find it.

Identify the bottleneck by seeing where you have the greatest upstream inventory piling up, and...

...

• Working on parts you don’t need urgently diverts capacity from the more important backlog.
• Occasionally, the bottleneck doesn’t have the upstream inventory to work at full capacity (and technically isn’t the bottleneck at that moment).
  • This may be because the non-bottlenecks are working on non-bottleneck parts or haven’t built up inventory.
• Machines run idle because people are redistributed to work on non-bottlenecks.

Fixes to Bottlenecks

We can divide this into a few themes.

Improve the Bottleneck Itself

• Skip unnecessary steps or decreasing setup/switching costs.
• Add on supplements to increase bottleneck capacity, even if they’re less efficient
• Guarantee round-the-clock production at the bottleneck.
  • Example: Eliminate lunch breaks and downtime.
• Permanently staff people at bottlenecks to decrease idle time.
  • Eg have people waiting by dishwasher to prepare loads and unload immediately.
  • Remember that the cost of a lost hour at this bottleneck is very expensive, and possibly well worth people idling on standby.

Improve What the Bottleneck is Working On

• Focus the bottleneck only on parts that are...

...

Buffer

The bottleneck should have surplus inventory upstream so it doesn’t idle.

Analogy: non-bottleneck boys can scout ahead, clear brush so the bottleneck boy can keep walking at normal pace without having to stop.

This buffer allows for “good enough” scheduling rather than needing to be perfectly accurate.

Goldratt suggests choosing a time buffer equal to half the current lead time, then decreasing or increasing as deadlines or hit or missed.

Rope

When non-bottlenecks exceed a certain surplus level, they idle.

Analogy: tie a rope between the boy in the front and the bottleneck boy, and limit the maximum distance between the two.

Similarly, prevent work-in-process inventory from exceeding a threshold level.

• Henry Ford purposely limited space for inventory to detect bottlenecks.
• In Toyota manufacturing, inventory is limited to containers containing a number of units, marked by a card. When this container is withdrawn for further processing, the card is returned to the upstream work center - only then can the center produce.
• In Kanban software engineering, no work can be added to a pipeline until the existing work has been moved to the...

...

Distilling Facts into Principles

Collecting too much information without identifying the underlying intrinsic order leads to false patterns and bad decision making.

Before the periodic table, it was unclear how one should understand the chemical elements. By color? State of matter? Instead, Mendeleev organized the elements first by atomic weight, then by reactivity (eg sodium and potassium behave similarly when thrown in water). This even allowed prediction of elements that didn’t yet exist.

To find the underlying intrinsic order, start from simpler If-Then principles. Only then can you question, “if this is true, what can I predict to be true?” Then test these predictions and continue testing the hypothesis. (Shortform note: This section felt a bit shoehorned in by Goldratt as a pedagogical item.)

...

But after identifying bottlenecks, it’s not clear how to increase capacity. They realize the bottleneck machine is idling at times because of union rules; also, the heat treatment inefficiently runs small loads.

They reorder the work queue so that the bottleneck is working only on the oldest unshipped orders. They clear the idle time with the union rep.

They run into an issue where the bottleneck doesn’t have upstream parts available. They learn the non-bottlenecks are working on non-bottleneck parts. They put red tags on bottleneck parts so that those always get highest priority.

Making Progress

They start filling their backlog and reduce their lead time. But it’s not enough to clear the backlog entirely.

• They add an extra outdated machine that will increase capacity at the bottleneck, albeit inefficiently.
• They start collecting better data about the performance at their bottlenecks.
• They load and unload at heat treatment immediately, letting workers idle there instead of letting the bottleneck idle.

New bottlenecks seem to emerge. It turns out they’re releasing too much upstream material and focusing on producing red bottleneck parts only, leading...