Hello TOC

Finance & Measurements, as Taught by Eli Goldratt · Lesson 3 of 8

Justifying the machine

A $100,000 machine passes through five desks and comes out with a one-year payback. Follow each step of the standard calculation, then see what really happens if the machine is a non-bottleneck, and what the same calculation does to a machine that is the bottleneck.

18 min · Not completed yet

An appropriation request lands on Helen’s desk: a new $100,000 machine that makes parts twice as fast as the old one. It has already passed through four desks. Each one added a number.

Five desks, one number

Here is how the request was built. Notice what each person had to assume.

  1. The engineer calculates the time saved per part. “There are so many parts this machine can make. I’ll have to use an average.” Time saved per part: 5 minutes.
  2. The materials manager estimates how many parts it will make a year. “I’ll take the forecast as a base and make some calculations.” 30,000 parts a year, so 2,500 hours saved a year.
  3. The cost accountant turns hours into money. “This department’s overhead factor is 1.83, the plant’s 1.36, the division’s 1.57. Should I add corporate overhead too?” Direct labour $8 an hour, overhead factor 4: $8 + 4 × $8 = $40 an hour, so $100,000 saved a year.
  4. The comptroller checks it. “This return is too good. I’m sure they fudged the numbers.” Machine price $100,000: payback one year.
  5. The CEO compares it with the standard. Equipment must pay back within two years. Approved: at last, an easy decision.

Scenario one: the machine is a non-bottleneck

Most machines are non-bottlenecks, so this is the usual case. Will the investment really be returned in one year?

What will actually happen to the $100,000 of “savings”?

  • They will appear, as calculated
  • They will appear, but only after two years, once the operators are fully trained
  • They won’t materialise; inventory will rise; throughput may even suffer

Goldratt’s three reasons are below. There is no way the machine pays back in a year.

The savings won’t materialise. New equipment rarely reduces overhead. It usually doesn’t reduce even direct labour: firing people because of a better machine risks a revolt, and moving them to another department doesn’t reduce the cost.

Inventory will increase. To “justify” the investment, everyone will try to keep the new machine busy. It isn’t a bottleneck, so running it flat out builds a steady stream of surplus inventory.

Throughput may suffer. With the faster machine, the standard time for the parts is cut. Foremen still using the old machines are now condemned by their variance measurements: ten minutes spent on something “that should take five”. If the old machines together have much more capacity than the new one, work drifts to the new one until it becomes the bottleneck.

Scenario two: the machine is the bottleneck

There is no third scenario. Now suppose the machine is for the bottleneck, but it is only 10% faster, not twice as fast.

The standard process says: twice as fast gave a one-year payback, so 10% faster gives a ten-year payback. The investment is condemned.

Brightline’s sales are $10 million a year; raw material is about 50% of the selling price. The 10% faster machine adds about 5% capacity to the bottleneck, and the market would buy the extra output. What is the real payback?

  • Ten years, as the standard says
  • A few months
  • Never

On the bottleneck, the justification is added sales, not cost savings. 5% more capacity is 5% more sales: $500,000 a year. Half of that is raw material, so profit rises by $250,000 a year on a $100,000 machine: paid back in about five months.

What the two scenarios show

The standard judgment does two kinds of damage at once:

  • It approves investments in non-bottlenecks, whose savings never appear and which create inventory.
  • It condemns investments in the bottleneck, whose real payback is measured in months.

The calculation is careful, detailed and audited. It is also aimed at the wrong thing: local cost instead of the company’s throughput.

Check your understanding

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

1. The cost accountant priced each saved hour at $40 because…

  • That is the operator’s wage
  • The machine costs an hour to run, including power
  • The market rate is $40
  • Labour of $8 was loaded with an overhead factor of 4

8 + 4 × 8 = 40.

2. For a non-bottleneck machine, the predicted savings usually…

  • Don’t materialise: overhead and labour don’t really fall
  • Appear in full
  • Double
  • Appear in full, but only after two years of running

Unless someone actually leaves the payroll, nothing is saved.

3. Trying to “justify” a non-bottleneck machine by running it flat out leads to…

  • Higher throughput
  • Lower lead times
  • Surplus inventory
  • Lower overhead

Activating isn’t utilising.

4. For a bottleneck, an investment should be justified by…

  • Cost savings
  • Added sales and the throughput they bring
  • Machine speed
  • The standard two-year payback rule applied to savings

An hour gained at the bottleneck is an hour gained for the system.

5. The 10% faster bottleneck machine actually paid back in about…

  • Ten years
  • Two years
  • Never
  • Five months

$250,000 a year of extra profit on $100,000.