Book notes: The Goal


17 January 2025

After years of hearing about it from my process improvement nerd friends, I finally read “The Goal” by Eli Goldratt. It became something of a cult classic in the 80s. It’s 40+ years old now but still feels like a useful “business novel” to introduce the Theory of Constraints (ToC). ToC is a management philosophy aimed at improving organisational efficiency by identifying, and carefully leveraging bottlenecks to increase overall throughput.

The story talks about the challenges of Alex Rogo, a manager at an industrial plant somewhere in middle America in the 1980s struggling to help his failing plant to meet their deadlines and regain profitability. I, myself, grew up in the Midwest in the 1980s so the setting gives me a twinge of nostalgia. More importantly, the story is full of great lessons about maximising throughput — not efficiency — and helping a team to succeed at achieving its one most important goal.

Here are my key takeaways:

Key concepts

Throughput: revenue/time

This isn’t just how much you produce or even how quickly you produce it, but how quickly you sell your product. Think of it as revenue velocity. This is a subtle but hugely important point that often gets overlooked by efficiency geeks like me who start chasing “local maxima” and sometimes miss the bigger picture. It doesn’t matter how quickly you product things if you’re not also selling them just as fast.

Inventory: the value of things unsold

Nearly every part of every production systems carries some inventory, these are things we’ve bought or invested in but haven’t yet sold. Things that are partially finished but haven’t yet made it into our customer’s hands – whether they’re internal or external. Consider work in progress (WIP) or even tickets in a backlog that we’ve invested in refining but will never actually deliver because other work will emerge as more valuable over time.

Operational expense: investment/throughput

Think about all the money we put into the system in order to increase our throughput (our rate of making money): Our salaries, the rent on our buildings, electricity, insurance, services provided by vendors, advertising, and especially the carrying costs of our inventory. Every penny we spend trying to make money is operational expense.

Each of these is important but the one true goal, according to the story, is to “increase throughput while simultaneously reducing inventory and operating expense”.

What gets in the way?

There are two major gotchas that prevent us from achieving “The Goal”.

Dependent events

Certain things have to occur in sequence. For example, you can’t build a sub-assembly until you have each of the parts. You can’t perform quality control on the entire product until the whole thing is complete. Understanding the immutable flow of dependent events is a key discipline. As is identifying which events are not dependent on each other. We sometimes assume that B must follow A but it isn’t always the case and bears close examination.

Statistical variation

Whether we’re making toasters or SaaS products, most manual production operations are “stochastic” in nature meaning that they can be statistically analysed but not accurately predicted in isolation. If you want an easy bet, ask a friend how many heads or tails you’ll get from 10 coin tosses.

Subtly point out that the statistical odds of getting a heads or tails in any coin toss is 50%. Your friend will likely say 5 out of 10 coin tosses should be heads. Statistically, this is true but in reality, a single set of 10 coin tosses is more likely to produce something else. Maybe 4 heads and 6 tails, or 7 heads and 3 tails, or even 9 heads and 1 tails but probably not exactly 5 heads and 5 tails.

In our production line this means we can’t perfectly predict how long a process will take or – put another way – exactly how many units we can produce in a certain time period. If downstream capacity is fixed at a certain rate, then trying to “catch up” by producing more units in less time and delivering a larger batch size won’t necessarily help and could even introduce greater bottle necks as the downstream system struggles to catch-up to the increasing backlog.

Simulations & games

Hiking with Scouts

At one point in the book, Alex Rogo, the protagonist, goes hiking with a group of Boy Scouts. The path is narrow and each person can only go as fast as the person in front of them, but they can always go slower making each person in front the bottleneck. Slowness accumulates throughout the system and the “product” isn’t complete until every hiker gets to the end. If the faster people are are at the front, the column of people just gets longer but it still takes as long as it takes the slowest person to walk the whole way. It’s possible to periodically rush everybody to catch up but this burns extra energy.

In this example, inventory is the length of the column. Operational expense is the energy consumed and throughput is the rate at which the last person walks the trail. The group consumes the “raw materials” of the un-walked trail and the rate of demand is how many miles per hour they’d need to walk to get to the campsite by nightfall.

Matches and bowls

On the hike, Alex gets an idea for a game to illustrate dependent events and statistical variation.

Here’s how it works.

Materials

  • A box of matches or other small objects
  • A 6-sided die
  • A number of small bowls, boxes, or containers
  • A player for each bowl

Playing

Line the bowls up next to each other with the box of matches at the left end.

Each player rolls the die, starting at the left, and moves that many matches into the next player’s bowl but you can only move as many matches as you have in your own bowl. So if you roll a 5 but you only have 3 matches, you can only move 3 matches to the bowl on your right.

A cycle is defined as every player rolling the die once.

Ask the group, how many matches they think they can move to the final bowl after a single cycle. The average roll of a 6 sided die is 3.5 because there’s no “zero”. So someone is likely to say 3.5 matches. And after 10 rounds, you should have 35 matches in the last bowl, right? Wrong. Dependent events and statistical variation mean that smaller upstream rolls of the die create growing shortages downstream and even if the downstream players roll higher numbers, they don’t have the inventory to move that many matches.

Try it and you’ll see that the actual number over 10 cycles is actually much lower than the expected “average” of 35.

Here’s a nice write-up and even a Google sheet simulation.

Bottlenecks & non-bottlenecks

A bottleneck is any operation or step in a process where the rate of demand exceeds the rate of delivery. Unprocessed inventory tends to pile up in front of the bottleneck. A non-bottleneck is an operation where the rate of delivery exceeds demand. Bottlenecks aren’t necessarily a bad thing, you just have to identify and position them in your process so they don’t slow down throughput or increase inventory or operational expense too much.

Conclusions

The rest of the book goes into an exploration of continuous improvement which was revolutionary in the 80s but feel pretty well covered and extended in more recent writing. But the basic ideas of throughput, inventory, and operational expense feel somewhat unique and quite relevant to any team or company. I enjoyed the “hiking” metaphor as a way to illustrate how bottlenecks constrain a system and why the order of those bottlenecks is significant. The language of “dependent events” and “statisical variation” feels useful to explain why it’s so difficult to predict with absolute certainty how long it will take for a complex system to produce results.

I’m pleased to put this one in my collection along side the Phoenix Project. Next, maybe I’ll finally read the The Unicorn Project.