Hello TOC

Operations, as Taught by Eli Goldratt · Lesson 7 of 8

Buffer management

DBR is the plan; buffer management is the execution. Divide the buffer into three regions, expedite only the red, aim to expedite about 5% of the work, and use the record of “holes” to focus improvement, avoid the trap of self-defeating success and see the bottleneck move before it moves.

22 min · Not completed yet

Six weeks into DBR at Northgate. Lead time has halved. But the questions keep coming to Daniel’s door.

Shearing foreman: “I’ve got 30 units of one job and 10 of another. Which first?”

Maintenance: “Three machines are down. Everyone says theirs is the most urgent.”

Expeditor: “When should I chase a job, and how hard?”

Ruth: “One mechanism answers all of these.”

The open questions

DBR gives the plan. Four questions remain:

  1. How should a non-bottleneck prioritise its tasks? A resource upstream of the bottleneck should do first what the drum needs first. A resource downstream should do first what the market needs first. Either way, it needs a mechanism to know.
  2. How should support functions prioritise? Maintenance, transport and tool rooms don’t work on the product but keep the flow alive. Today they prioritise by the level of screaming and personal relationships. Asking the foremen doesn’t help: they’re conditioned to ask for first priority. Their priorities should come from the drum.
  3. When should work be expedited? For 100% safety the buffer would have to be infinite. Expediting lets us get high protection from a reasonable buffer.
  4. How much should we expedite, and when should the buffer change?

The three regions

Work is released one buffer time before the bottleneck needs it, but the actual process time is much shorter than the buffer. So divide the buffer into thirds:

Region Time since release Status of a job not yet in the bank
I first third Not expected to be at the bank yet
II second third A fair chance it would be there
III (red) last third Expected to be there already

Buffer management combines the drum with the transactions reported on the floor to show, at any moment, where every job stands.

  • When to expedite: when a job is in region III, the red zone, and still isn’t at the bank. It should be there, and it isn’t.
  • How much to expedite: about 5% of the work. Much less than 5% means the buffer is too big: you’re wasting lead time and inventory. Much more means the buffer is too small: you’re wasting effort and putting throughput at risk.
  • How to prioritise: buffer management shows the urgency of every job queuing at every work centre. Non-bottlenecks work on the most urgent job first. Support functions serve the work centre holding the most urgent job.

A disturbance hits the shearing line. The planner wants to rerun the schedule for the whole mill tonight. What does buffer management say?

  • Good idea: an updated schedule always reflects the latest reality better than an old one
  • Don’t reschedule: change priorities through buffer status
  • Stop the bottleneck until the shearing line recovers

Rescheduling spreads a local disturbance across the whole system and amplifies it. The right split: DBR for planning, buffer management for execution. Goldratt suggested a simple test for scheduling software: ask how it deals with disturbances. A good answer talks about this separation; a bad one boasts about rescheduling everything in minutes.

The next leap: self-defeating success

A good DBR implementation cuts lead times and improves on-time delivery. A market that gets much better service buys more. But even a modest increase in sales eats a big share of the protective capacity. With less protective capacity, holes appear in the bank; due-date performance slips. And a market that has been spoiled by excellent service overreacts to the slip. The result can be a net loss of sales.

What assumption makes this success self-defeating?

  • That sales should not grow
  • That protective capacity can’t change
  • That the buffers should not be increased

Less protective capacity causes holes only if the buffer stays the same. Increase the buffer, and the self-defeat is prevented. The goal of an organisation is not zero inventory: the damage of missing due dates dwarfs the damage of a modest increase in inventory. Watch the expediting rate, and when it climbs to about 10% of the work, increase the buffer.

Focusing improvement: TOC and Lean

Expedite only when a hole has penetrated region III, and record where the job was stuck: which department, work centre or vendor. At the end of the month, add it up. The place that caused the most holes is where local improvement, Lean, Six Sigma, set-up reduction, should be focused. TOC and Lean aren’t rivals: TOC improves the whole process and tells Lean where local improvement will immediately become global improvement.

When the bottleneck moves

DBR orients the entire operation to the constraint. If the constraint shifts, everyone’s priorities must change, which can be costly. Buffer management helps in two ways.

It warns of a shift caused by a change in mix. Departments A and B each have 2,400 minutes a week. Orders come for 200 units of one product and 150 of another; B needs 2,600 minutes against 2,400, so B is the constraint. When the mix changes, the holes start appearing at a different resource before anyone has redone the calculation.

It shows where to add capacity so the bottleneck doesn’t move. Try this one:

The market orders 20 units a day. Red can process 22 a day, Green 18. Sales grow to 24 a day, and Green’s capacity is increased by 30%. Where is the constraint now?

  • Still Green
  • Red
  • The market

Green now does about 23.4 a day; Red still does 22. Red has become the constraint, and the sales increase is capped at 22. When you add capacity to the bottleneck, add the same percentage to the resources that buffer management shows causing most of the holes in region III, so they don’t turn into the next bottleneck.

Four benefits of buffer management

  1. Execution: it sets everyone’s priorities on line, enabling much smaller buffers: shorter lead times, less inventory, more throughput, better due dates.
  2. Focus: it points local improvement efforts where they become global results.
  3. Early warning: it signals a coming bottleneck shift due to mix changes.
  4. Growth: it shows where capacity must be added to take advantage of rising sales.

Goldratt added an encouraging point: implementing DBR needs a culture change, but putting buffer management in place does not, and yet it doubles the results.

Check your understanding

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

1. A job should be expedited when…

  • A customer calls asking where the order is
  • It is released
  • It is in region I
  • It is in region III and not yet at the bank

It should be there, and it isn’t.

2. If you expedite far less than 5% of the work…

  • The buffer is too big
  • The buffer is too small
  • Everything is perfect
  • The drum is wrong

You’re wasting lead time and inventory.

3. Disturbances should be handled by…

  • Rescheduling the whole plant from the new situation
  • Overtime everywhere
  • Changing priorities through buffer management
  • Stopping release

Rescheduling amplifies disturbances.

4. Self-defeating success is prevented by…

  • Refusing new orders
  • Increasing the buffers when expediting reaches about 10%
  • Raising prices
  • Cutting inventory further so that problems become visible

Zero inventory is not the goal.

5. Lean efforts should be focused on…

  • Every department equally, so nobody feels left out
  • The newest machines
  • The bottleneck only
  • The resources that cause the most holes in region III

That’s where local improvement becomes global.