Operations

Busy machines or on-time orders? How Drum-Buffer-Rope ends the trade-off

Mohit Saini · 23 Jun 2019 · 6 min read

Busy machines or on-time orders? How Drum-Buffer-Rope ends the trade-off

Every production planner lives with one dilemma, every single day:

  • Load the plant fully, and output looks good, until one breakdown or late supplier delays a chain of orders.
  • Keep some capacity spare, and orders are delivered on time, but on the days when nothing goes wrong, that spare capacity seems wasted.

For decades, manufacturing has treated this as a trade-off: high output or high reliability, pick one. Most planners try to walk a tightrope between the two.

This article traces how that dilemma has been tackled, and why Theory of Constraints offers a way to stop walking the tightrope altogether.

Attempt 1: calculate capacity precisely

The first family of answers assumed that if we could calculate capacity precisely enough, we could load the plant exactly to its limit.

  • MRP planned material backwards from due dates using fixed lead times, but assumed unlimited capacity.
  • MRP II added rough capacity checks. Plants still saw overloads and delays.
  • Monthly time buckets were introduced: plan a month, then re-plan for the next month to absorb spill-overs and forecast errors. This gave some stability within the month, but created a new pattern: a slow first week while plans are “netted” and suppliers catch up, and a frantic last week to hit the numbers.
  • Finite-capacity schedulers (APS and similar tools) used computing power to schedule every work centre and reschedule instantly when something went wrong. Every reschedule, however, moved someone’s promised date.

All these approaches share an assumption: capacity can be defined precisely enough to plan around. In reality, it can’t. Capacity changes with the product mix, with operator attendance, with humidity, with material quality, and with the many informal rules supervisors follow (like running light paint shades before dark ones) that they also break when an urgent order arrives.

Attempt 2: separate planning from execution

Dr. Eliyahu M. Goldratt took a different route. Instead of building a more detailed plan that would be broken by the first surprise, he proposed:

  • a simple plan focused on one resource, the constraint, and on flow, and
  • a robust execution system that sets priorities as reality unfolds.

This is Drum-Buffer-Rope (DBR):

  • Drum: the schedule of the constraint sets the pace for the whole plant, because the plant can’t produce more than its weakest link.
  • Buffer: a time cushion before the constraint and before shipping, so that normal disruptions don’t delay orders.
  • Rope: material is released only a buffer’s length before it is needed, which prevents the floor from flooding with work.

Day to day, buffer management tells everyone what to work on: orders are coloured green, yellow or red by how much of their buffer has been used.

A plant on one page

Picture a plant that makes pumps in four steps: casting, machining, assembly and testing. Machining can handle 40 pumps a day; every other step can do 55 or more. Machining is the constraint.

  • Drum: the plant can ship 40 pumps a day and no more, however hard the other steps work. So machining’s schedule becomes the plan for the whole plant, with orders sequenced on it by due date.
  • Rope: casting may start an order only a fixed time, say four days, before machining is due to take it up. Casting could produce 55 a day, but it produces only what the drum will consume. Its extra capacity is not used to build piles.
  • Buffer: those four days are the protection. If casting has a breakdown, machining still has work waiting. After machining, assembly and testing have enough spare capacity to catch up on their own.

Now look at what changes on the floor. Before, casting ran flat out, and castings piled up in front of machining: three weeks of work in progress, with the supervisor picking whatever was easiest to set up. After, there are about four days of work in front of machining, all of it in due-date order. Lead time drops from weeks to days, because orders no longer wait in a queue that was never needed.

Casting is now idle for part of the day. On an efficiency report, that looks like waste. On the plant’s output, it costs nothing, because machining, which sets the output, is never starved.

Learning from practice: simplified DBR

As DBR was implemented widely, practitioners noticed some problems:

  • In most plants, the market is the real constraint, not a machine. Demand swings, so the internal bottleneck is busy in some months and idle in others.
  • Following a detailed constraint schedule at all costs often clashed with customer due dates, which matter more.
  • Managing several buffers at once was complicated, and managers trying to squeeze every minute out of the constraint loaded it so far ahead that delivery suffered.

Simplified DBR (S-DBR), described by Eli Schragenheim and H. William Dettmer, addressed this:

  • The drum becomes the market’s demand: firm orders and their due dates.
  • The busiest internal resource is treated as a capacity-constrained resource (CCR) and loaded only to a planned level, say about 90%, leaving protection for surprises.
  • Only one buffer remains: the shipping buffer, from material release to dispatch.
  • Material is never released before its release date.
  • New orders get due dates based on the planned load on the CCR, so promises reflect reality.

Make-to-availability

Original DBR treated stock items as if they were orders with artificial due dates. That made priorities meaningless: a “late” stock order might be for an item that was plentiful in the warehouse.

For stocked items, TOC switched to stock buffers: each item has a target stock, production replenishes what was consumed, and priority depends on how far the item’s stock has fallen below target. Goldratt called this make-to-availability: a promise that the item will be on the shelf, not just that it was made.

Safe dates and dynamic release

Two further refinements came from practice:

  • Safe dates: when operations tells sales which delivery dates it can safely promise, based on current load, the company can quote short, reliable dates and even encourage orders in quieter periods to smooth the load.
  • Dynamic release: instead of releasing material on a fixed date regardless of conditions, the release is adjusted to keep work in progress roughly constant. When the floor moves faster, more is released; when there is a disruption, release slows. This borrows a strength of pull systems (limiting work in progress) while keeping DBR’s due-date logic.

What this means for the planner’s dilemma

Look back at the original trade-off. It assumed that spare capacity is wasted and that a fully loaded plant gives the most output. DBR breaks both assumptions:

  • Protective capacity at non-constraints isn’t waste. It is what allows the constraint and the due dates to be protected.
  • Releasing less work gives more output. With less work in progress, lead times shrink, priorities are clear, and the constraint is fed with the right work.

The goal is not to plan capacity perfectly. It is to plan simply and protect the plan with buffers.

Plants that make this shift typically see lead times fall sharply, on-time delivery rise to well above 90%, and output go up, from the same people and machines.

Mohit Saini

Strategy, supply chain & sales transformation consultant. About · Book a discussion