Bottlenecks & debottlenecking

A process can only go as fast as its slowest step. Picture a factory where each product passes through four stages:

Stage	Capacity per month
Cutting	10,000
Assembly	8,000
Painting	12,000
Packaging	15,000

The plant can’t ship more than 8,000 a month, because assembly is the bottleneck. Buying a faster cutting machine changes nothing except the pile of half-finished parts in front of assembly. An hour lost at the bottleneck is an hour lost for the whole plant; an hour saved anywhere else saves nothing.

Eliyahu Goldratt’s theory of constraints (TOC) turns this into five focusing steps:

1. Identify the constraint.
2. Exploit it: never let it stand idle. Run it through breaks, keep a small buffer in front of it, inspect before it so it never works on bad parts.
3. Subordinate everything else to it: other stages work at its pace, not their own.
4. Elevate it: only now spend money, on an extra shift, a second machine or outsourcing.
5. Repeat: once assembly is faster, something else becomes the constraint.

Steps 2 and 3 are often free and can unlock a surprising amount of output before any capital is spent.

Finding the constraint on a real floor

• Load against capacity: compute each resource’s utilization for the plan; the highest is the likely bottleneck.
• Where work piles up: queues of work in process build in front of the constraint, while the stages after it often wait.
• Starved and blocked time: stages downstream of the bottleneck sit idle for lack of work; stages upstream stop because there is nowhere to put their output.
• The tell-tale question: if this machine got one more hour, would the plant ship more?

What an hour is worth

Assembly makes 8,000 units a month in 200 hours, 40 units an hour. Each unit earns ₹500 of throughput (selling price minus truly variable costs such as materials).

• An hour lost at assembly loses 40 units: ₹20,000 of throughput, gone for good.
• An hour lost at cutting, which has spare capacity, loses nothing, as long as assembly is never starved.
• If 5% of assembled units are scrapped later because of defective cut parts, inspecting before assembly frees 5% of its time: 400 more good units a month, worth ₹2 lakh.

Choosing the mix at the constraint

When the bottleneck can’t make everything demanded, rank products by throughput per bottleneck minute, not per unit. Product P earns ₹600 a unit but uses 15 bottleneck minutes (₹40 a minute); product Q earns ₹400 but uses 5 minutes (₹80 a minute). Q should get priority, even though P looks more profitable per unit.

Drum, buffer, rope

The theory of constraints schedules the whole plant around the bottleneck:

• Drum: the bottleneck’s schedule sets the pace for everything.
• Buffer: a time buffer of work in front of the bottleneck protects it from upstream hiccups, so it never starves.
• Rope: materials are released into the plant only as fast as the drum consumes them, which stops work piling up everywhere else.

Balanced lines lose output

It seems ideal to give every stage the same capacity. But stages depend on each other, and each one varies from hour to hour. When a stage has a slow spell, the next one waits; when it has a fast spell, it can’t run ahead of the one before. With no spare capacity anywhere, losses accumulate and nothing can catch up. Goldratt illustrated this with a dice game; queueing theory confirms it. Deliberately unbalanced lines, with one clear bottleneck and spare capacity elsewhere, flow better.

Constraints move and hide

• Mix-dependent bottlenecks: one product mix overloads assembly, another overloads painting. The constraint can wander from week to week.
• Policy constraints: often the limit is a rule, not a machine: a one-shift policy, a minimum batch size, a quality release that happens only once a day.
• Outside the plant: a supplier’s capacity, a port, warehouse dock doors or a transport fleet can be the real constraint.
• The market: when capacity exceeds demand, the constraint is sales. Making more only builds stock; the right move is to sell more or cut capacity costs.

Batch sizes at and away from the bottleneck

At the bottleneck, every changeover steals output, so running larger batches there can be worth the extra inventory. Away from the bottleneck, changeovers use spare time, so smaller batches cost little and keep inventory and lead times down. Splitting a production batch into smaller transfer batches, moving the first parts on before the whole batch is finished, also cuts lead time without adding changeovers.

Measuring it right

Local efficiency targets (“keep every machine busy”) work against flow: they reward non-bottlenecks for producing work that only queues in front of the constraint. The theory of constraints replaces them with throughput (money coming in through sales), inventory and operating expense, judged for the whole system.