The best-optimized plant is almost always the one that ships nothing
Eliyahu Goldratt published a novel, not an essay, built around a plant manager 90 days from bankruptcy. Forty years later, The Goal is still one of the most useful reads for anyone turning around a struggling industrial site. By Mounir Telkass, founder of MT-Transition.
Goldratt’s Theory of Constraints: one single station sets the pace for the entire plant
Eliyahu Goldratt was an Israeli physicist. In 1984 he published a novel — not an essay, a novel — built around an American plant manager 90 days from bankruptcy. The Goal tells the story of how he saves his site. Forty years later, it is still one of the most widely assigned reads in engineering schools and industrial management circles. What Goldratt managed to formulate, before anyone else: most decisions made on a shop floor are made using the wrong metrics.
The consequence — and this is the substance of the book — is that plants methodically optimize the wrong stations. Utilization rates look flattering on paper. And the plant as a whole ships no more product than before. Sometimes less. The Theory of Constraints (TOC) he proposes as the way out of this trap comes down to a few simple principles — and it’s precisely because they’re simple that they still unsettle industrial leadership teams who have spent twenty years measuring the wrong thing.
Three mechanisms from the book deserve close reading by anyone running an industrial site.
1. One single station sets the throughput of the entire plant. Find it.
In any production line, a single station determines the output of the whole. That’s the bottleneck. Not the most visible one. Not the most expensive one. Not the one you’d guess. But it’s the one that decides how much the plant ships per day. Every other station can run at full capacity: if the bottleneck keeps pace, fine. If it slows down, the entire plant slows with it — whatever the other stations are doing.
The five focusing steps for managing by the constraint
1. Identify the bottleneck (often invisible in standard KPIs).
2. Exploit it to the maximum — never a second of downtime, staggered breaks, never a stop for a minor incident.
3. Subordinate everything else to its pace. Non-bottlenecks work to the bottleneck’s tempo, not their own.
4. Elevate the bottleneck — targeted investment: hiring, capacity, method.
5. Repeat. The bottleneck has moved. Back to step 1.
This is a total inversion of classic industrial thinking. You stop trying to optimize locally. You accept that 80% of the plant may be “under-utilized” — because you’ve understood there’s no point running faster anywhere except at the bottleneck. For a site director taking on a turnaround mission, this is the first lens to apply before any Capex plan. In most cases, the planned investment won’t fix the bottleneck — it will actually make it worse, by pushing even more flow toward an unchanged constraint.
2. Three metrics are enough. All the others lie.
Goldratt dismisses classic industrial cost accounting in two pages. Unit cost, machine utilization rate, output per operator — all these indicators are, he argues, structurally misleading in a modern plant. They measure local activity, not overall performance. Instead, three metrics, and three only:
Throughput: the rate at which the system generates money through sales. Not production. Sales.
Inventory: the money tied up in everything the company has bought that hasn’t yet been sold — raw materials, work in progress, finished goods.
Operating expense: the money spent turning inventory into throughput.
A decision is good if it increases Throughput, and/or reduces Inventory, and/or reduces Operating expense. Otherwise, it isn’t. That’s the whole test.
Applied to real plants, this is unforgiving. A workshop “90% busy” producing stock nobody ordered improves its local metric — and simultaneously worsens all three of Goldratt’s measures. Inventory rises. Throughput is unchanged (nothing extra gets sold). Expenses rise (energy, handling, storage space). This is the lens to impose in the executive committee before any investment, any hiring, any improvement project. If a decision doesn’t move one of the three metrics in the right direction, it isn’t an improvement — it’s agitation. And agitation is expensive.
3. An hour lost at the bottleneck is an hour lost for the whole plant
“An hour lost at the bottleneck is an hour lost for the entire system. An hour saved at a non-bottleneck is a mirage.”
This is the simplest expression of the whole theory. And the ultimate test for any operational decision. At the bottleneck, every minute counts absolutely. A 20-minute machine stoppage at the bottleneck station is 20 minutes of throughput lost for the entire plant — for good. Not 20 “local” minutes. 20 minutes of revenue. Permanently.
Conversely, at a non-bottleneck station, a 30% productivity improvement changes nothing about the plant’s throughput. The station simply finishes its batch earlier, then waits for the bottleneck. Invisible in local KPIs. Brutal in the accounts. A continuous-improvement team spending six months on a non-bottleneck station is six months of effort that will never earn the company a single euro.
A mission that chains together all three principles
Mid-sized food processing company, 280 employees, two parallel lines. Margins eroding for three years. The outgoing plant director had launched a major plan: Capex for a new packaging line, hiring of 18 FTEs, a Lean program on the mixing workshop. Result 18 months later: identical throughput, doubled inventory, EBITDA still declining. An transition industrial director arrives. No grand plan. One week walking the floor with a notebook and a stopwatch.
The “identify” step
The real bottleneck isn’t the new packaging line (running at only 60%). It isn’t the mixing workshop either. It’s a quality-control station after baking, where only one trained operator can sign off on a batch. When that person is on leave or a long break, the entire plant slows down. Nobody had spotted it — because this station showed a 75% utilization rate, which looked reasonable.
The “exploit” step
The station is moved to staggered breaks, a second sign-off operator is trained within 6 weeks (training that could have happened 4 years earlier), non-critical checks are shifted upstream to non-qualified technicians ahead of the final sign-off. Bottleneck throughput: +25% in 8 weeks, without spending a euro.
The “subordinate and verify the three metrics” step
The transition director freezes the Lean program targeting the mixing workshop. He reassigns the continuous-improvement teams to the quality-control station. He officially accepts that the new line runs at 60% — because it sits downstream of the bottleneck, and running it faster would only build inventory.
Six months later: Throughput up 25%. Inventory down 15%. Expenses down 8%. All three metrics move in the right direction at once. That’s the Goldratt signal. At 12 months, EBITDA is fully recovered. As for the new packaging line, it probably shouldn’t have been bought in the first place — but that’s another story.
What to take away
1. The entire plant runs at the pace of a single station — and it’s almost never the one you’d guess. Find it before investing a euro.
2. Three metrics are enough: Throughput, Inventory, Operating expense. Everything else is local, and therefore misleading.
3. An hour lost at the bottleneck equals revenue lost for good. An hour gained elsewhere is a mirage. Prioritize budgets accordingly.
The book is 40 years old. Most European plants still measure performance where it isn’t. That’s precisely what keeps it worth reading — and no site director should end their career without having read it twice.
Mounir Telkass — MT-Transition, industrial transition management firm. See also our profile on the transition industrial director role.
A similar situation at your company?
Callback within 2 business hours. You’ll speak with an industry expert, not a salesperson.
☎ Call — +33 6 59 15 73 54Callback within 2 business hours · 3 targeted profiles within 72h · 100% industry