Hello TOC

Operations, as Taught by Eli Goldratt · Lesson 4 of 8

Is an idle resource a waste?

Goldratt's proof, using only three facts true of every operation: try to keep even one resource busy all the time and see what the others must do. Then test yourself on four decisions where the old paradigm and the new one give opposite answers.

22 min · Not completed yet

Ruth: “Viktor, I want to prove to you that it pays to stop most of your resources, deliberately, from time to time.”

Viktor: “Stop them? In the name of productivity?”

Ruth: “Exactly in the name of productivity. Give me three facts about your mill, and I’ll do the rest.”

Three facts about any operation

Goldratt didn’t prove his point on a special case. He used three characteristics that fit almost every operation. Check each one against your own.

  1. Reality is not simple. More than one resource is needed to deliver the product or service: a machine, a team, a department. (If one resource does everything, you don’t have an organisation.)
  2. Reality is not chaotic. Work doesn’t flow uniformly; it skips some resources and goes back and forth. Still, a general direction of flow can be identified.
  3. Reality is not predictable: Murphy exists. From time to time a disturbance makes one resource work slower than average. You can and should reduce Murphy, but you can’t eliminate it.

The experiment

If an idle resource is a major waste, we should strive to keep every resource working 100% of the time. Let’s see what happens when we try it for just one resource, any one. Call it Green.

Murphy exists, so the upstream resources will sometimes stop. The only way to keep Green working constantly is to build a bank of work in front of it. When Murphy strikes upstream, Green keeps going by eating into the bank. (The bank has a price: inventory and lead time in production, irritated customers in service.)

Now the disturbance is over. The upstream resources must do two things at once: keep feeding Green and rebuild the bank. So they must be able to produce more than Green: they need protective capacity. And they must rebuild the bank before Murphy strikes again, so they need a lot of it.

How much is “a lot”? Suppose the upstream resources have 5% more capacity than Green. When Murphy strikes, the bank drains at Green’s full rate. It refills only at the 5% surplus: twenty times slower. For all that time, Green is exposed; if Murphy strikes again, Green stops.

Pause and think: the same logic says the resources after Green need protective capacity too. Why?

  • They don’t; only upstream matters
  • If one stops, Green’s output piles up; to catch up while Green keeps going, it needs spare capacity
  • Because of the measurement system, which judges every resource on its own efficiency and output

Downstream, a stoppage creates a queue that must be cleared while new work keeps arriving at Green’s pace. Only spare capacity can clear it.

What if we keep all the others busy too?

We now know: to keep one resource at 100%, all the others must have significant protective capacity. So what happens if we also insist that they work as much as they can?

They have much more capacity than Green, so they will produce much more than Green can use. Mountains of inventory, longer lead times: major waste.

The conclusions, in Goldratt’s order:

  • Resources standing idle are not necessarily a waste.
  • Sometimes resources that are not standing idle are the major cause of waste.
  • To avoid waste, when the bank is full, stop the other resources from over-producing.

The assumption under the waste arrow is invalid. The cloud from the last lesson evaporates.

Activating is not utilising

Goldratt put the result in one distinction. Activating a resource means making it work. Utilising a resource means making money from its work. They are not the same thing.

  • When the constraint is outside operations (there aren’t enough orders), no resource can be utilised 100%. Activating them to 100% makes more inventory, not more money.
  • When the constraint is inside operations, we just proved that among resources making the same products there can be only one that is utilised 100%. The others can be activated, but it won’t make more money. There can’t be more than one bottleneck.

So the bottleneck’s utilisation dictates the throughput of the whole system:

An hour lost at a bottleneck is an hour lost for the entire system; an hour saved at a non-bottleneck is a mirage.

This is not a triviality. It is a major paradigm shift: the local impact of an hour is not the impact on the system. Test yourself.

Test yourself: four decisions

1. Counting depreciation, operating expense and overhead, a work centre costs $300 an hour. An hour is lost there. How big is the damage?

  • , since that is what the work centre costs per hour whether it runs or not
  • We can’t say until we know whether it is a bottleneck

$300? Sorry: that is the old paradigm. Can’t say yet? Good, you’ve started to adjust. If it is the bottleneck, the damage is the throughput per hour of the product lines that go through it, often far more than $300. If it is a non-bottleneck with enough protective capacity, the damage is zero.

2. Resource C is the bottleneck. Some of its work can be offloaded to B, which has plenty of capacity but is much less efficient: it takes twice as long. The offload needs a $5,000 investment and adds 5% capacity on C. Sales are $5 million a year; raw material is 50% of the price. What is the payback?

  • It isn’t worth it: B is half as efficient
  • About three years
  • About two weeks

Two weeks. 5% more on the bottleneck is 5% more sales: $250,000 a year. Raw material for those sales is $125,000. Operating expense doesn’t change: B’s people are already paid. So profit rises by $125,000 a year on a $5,000 investment, paid back in about two weeks. B’s “inefficiency” costs nothing, because B has time to spare. Work should be offloaded from the bottleneck; cost considerations are usually negligible beside the throughput gained.

3. Resource D is not the bottleneck and has ample protective capacity. The company has a no-layoffs policy. One proposal: a new $100,000 machine giving D 50% more capacity. Another: $20,000 of software giving D 20% more. What should you do?

  • Buy the machine
  • Buy the software: it’s cheaper
  • Don’t spend the money

Extra capacity at a non-bottleneck sells nothing more, and with no layoffs it saves nothing either. Paying to make a non-bottleneck more efficient is usually a waste.

4. The forklift driver loads the truck with everything else first, to save a set-up, and the bottleneck waits an hour for its material. Was saving the set-up worth it?

  • No: an hour lost at the bottleneck is an hour lost for the whole operation
  • Yes: a set-up always costs money, and the truck has to be loaded either way

Saving set-ups on a non-bottleneck can do considerable damage when it starves the bottleneck.

What must change on the shop floor

All resources should be managed so that the bottleneck is utilised as much as possible. Common habits that waste the bottleneck:

  • Non-bottlenecks work on other things when they should be working on what the bottleneck needs.
  • The bottleneck isn’t manned all the time: lunch breaks and shift changes stop it.
  • Tools, parts or set-up materials aren’t ready when it needs them.

And one habit that wastes money: buying efficiency for resources that aren’t the bottleneck.

Check your understanding

Answer all five. Get four right to complete the lesson.

1. To keep one resource working 100%, despite Murphy, we need…

  • Overtime on every resource, so that any loss can be made up the same day
  • Better forecasts
  • A bank of work in front of it, and protective capacity on the other resources
  • A new machine

The bank protects it; spare capacity rebuilds the bank.

2. With 5% protective capacity, rebuilding the bank takes…

  • About twenty times longer than draining it
  • About the same time as draining it, if the resources work at full speed
  • Twice as long
  • No time

100% ÷ 5% = 20.

3. Activating a resource and utilising it differ because…

  • Activation costs less
  • Utilisation is measured in hours worked, activation in units produced
  • They are the same
  • Utilising means making money from its work; activating may only make inventory

Only the bottleneck’s work turns directly into throughput.

4. An hour saved at a non-bottleneck is…

  • Worth its cost per hour
  • A mirage: the system’s output doesn’t change
  • Worth an hour at the bottleneck, since all hours cost the same
  • Lost for the whole system

The local impact is not the impact on the system.

5. The offload to a less efficient resource paid back in about…

  • Three years
  • Never
  • Two weeks
  • Six months

$125,000 a year of extra profit for $5,000.