Capacities from Worked Example 1: Op 10 35, Op 20 40, Op 30 22, Op 40 50, Op 50 30, Op 60 45 units/day. System output is the smallest of them, 22. Capacity sitting above that rate: (35−22)+(40−22)+(50−22)+(30−22)+(45−22) = 13+18+28+8+23 = 90 units/day that exists, is paid for, and can never ship. Panel 2 plots output = min(Op 30 capacity, 30), because 30 is the next-lowest capacity in the same table — so elevating Op 30 pays one-for-one only from 22 to 30. The entire prize at the constraint is 8 units/day; at the article's $1,200 per unit, $9,600 a day. Everything past 30 is Step 5: the constraint has moved to NDT.
The Five Focusing Steps — In Depth, for Aerospace
Eli Goldratt’s Theory of Constraints rests on one insight: every production system has exactly one constraint that limits its output. Not two. Not a list. One. Everything else in the facility is either serving the constraint or generating waste. The Five Focusing Steps are the method for systematically managing that constraint.
The reason this matters for the Process Architect is that it provides a decision framework: where should you invest your engineering time? The answer is always: at the constraint first. Every hour of improvement at the constraint translates directly to system throughput. Every hour of improvement at a non-constraint translates to exactly zero system throughput.
Step 1: IDENTIFY the Constraint
The constraint is the operation with the least capacity relative to demand. In theory, it is the resource with the longest queue. In practice, finding it is harder than it sounds because the apparent constraint and the real constraint are often different.
Three practical methods for identification:
Queue depth analysis: Walk the floor and measure WIP at every operation. The operation with the most WIP in front of it is likely the constraint. Be thorough — count units on machines, in queues, on hold, and in storage.
Utilization observation: The constraint is the resource that is always busy. Other machines have idle time; the constraint does not. If you can observe all resources simultaneously during a normal production day, the constraint reveals itself.
The sprint test: Temporarily add capacity at the suspected constraint (extra shift, overtime, parallel resource). If system throughput increases, you found the constraint. If throughput does not increase, the real constraint is elsewhere — keep looking.
Step 2: EXPLOIT the Constraint
Get maximum output from the constraint without spending capital. This is the most under-executed step because organizations jump to Step 4 (buying equipment). Exploitation is free or near-free:
- Never let it starve: Ensure material is always waiting. A 15-minute buffer of WIP in front of the constraint ensures it never stops for lack of input.
- Run through breaks: Rotate operators so the constraint machine operates during lunch and breaks.
- Reduce changeovers: Apply SMED to the constraint first. Every minute saved in setup is a minute of additional production capacity.
- Eliminate rework at the constraint: Inspect before the constraint so defective parts never consume constraint time.
- Pre-stage everything: Tooling, programs, fixtures, documentation — everything the constraint needs should be ready before the current job finishes.
- Prioritize maintenance: The constraint gets preventive maintenance priority over all other machines.
Step 3: SUBORDINATE Everything Else
This is the hardest step psychologically because it means non-constraint resources must deliberately operate below their maximum capacity. Non-constraints should run at the pace the constraint can absorb — no faster.
Why this creates friction: traditional management metrics reward high utilization at every machine. Running a non-constraint at 70% feels “wasteful.” But Kingman’s Equation proves that running non-constraints below capacity provides the capacity buffer that absorbs variability shocks. And producing more than the constraint can process just builds WIP — which Little’s Law proves increases lead time for every job in the system.
Subordination in practice means:
- Upstream operations release work at the constraint’s consumption rate (the “Rope”)
- Downstream operations are always ready to process constraint output immediately
- Quality checks upstream of the constraint catch defects before they waste constraint time
- Maintenance prioritizes the constraint over all other resources
- When a non-constraint has idle time, it does not over-produce — it uses the time for setup reduction, maintenance, cross-training, or 5S
Step 4: ELEVATE the Constraint
Only after Steps 2 and 3 are fully executed: invest to increase the constraint’s capacity. This may mean buying a second machine, adding a shift, outsourcing overflow, or redesigning the process. This is the capital step — it costs money, so it must be justified by evidence that Steps 2–3 have been exhausted.
Step 5: REPEAT — Do Not Let Inertia Become the Constraint
When you elevate the constraint, a new bottleneck emerges elsewhere. Go back to Step 1. The most common failure mode is continuing to optimize the old constraint after it has moved — pouring resources into a machine that is no longer the bottleneck while the new constraint goes unaddressed.
⚠️ The Five Steps Are Sequential
Organizations that jump to Step 4 (buying equipment) before completing Steps 2 and 3 have permanently wasted their capital investment. If the old constraint had 30% untapped capacity hidden in changeover waste, break starvation, and rework, buying a second machine doubles capacity that was only 70% utilized. You now have two machines at 35% utilization, the same throughput you could have achieved with exploitation alone, and a capital expenditure that cannot be recovered.
The Financial Math of Constraint Management
Scenario: A machining value stream with 6 operations. Data collected over one week:
| Operation | Capacity (units/day) | Avg. Queue Depth (units) | Utilization |
|---|---|---|---|
| Op 10: Rough Mill | 35 | 8 | 74% |
| Op 20: Heat Treat | 40 | 5 | 65% |
| Op 30: Finish Mill | 22 | 45 | 96% |
| Op 40: Deburr | 50 | 3 | 52% |
| Op 50: NDT | 30 | 12 | 87% |
| Op 60: Final Inspect | 45 | 4 | 58% |
Identification: Op 30 (Finish Mill) has the lowest capacity (22 units/day), the highest utilization (96%), and the deepest queue (45 units). This is the constraint.
System throughput: The entire 6-operation value stream can only produce 22 units/day — regardless of the fact that rough milling can do 35, heat treat can do 40, and deburr can do 50. The constraint determines system output.
Scenario: The finish mill (constraint) loses 3 hours due to a tooling issue. What is the impact?
| Parameter | Value |
|---|---|
| Constraint throughput rate | 22 units/day = 2.75 units/hour |
| Downtime | 3 hours |
| Units lost | 2.75 × 3 = 8.25 units |
| Finished goods value per unit | $1,200 |
| Revenue impact | 8.25 × $1,200 = $9,900 |
Critical insight: Those 8.25 units are gone forever. The constraint cannot “make them up” — it was already running at 96% utilization. There is no spare capacity to recover lost production. Those units will not ship this week. If the program is on a 4-week delivery cycle, this single 3-hour event has reduced the month’s output by 8.25 units — potentially 1.5% of monthly production from one tooling failure.
Compare to non-constraint downtime: If Op 40 (Deburr) loses 3 hours, zero units of system output are lost. Deburr runs at 52% utilization — it will catch up within 2 hours of resumed operation. Non-constraint downtime is not a system event. Constraint downtime is a system event.
This is why Goldratt said: “An hour lost at the constraint is an hour lost for the entire factory, forever.”
Drum-Buffer-Rope Scheduling
DBR is the scheduling method that translates TOC theory into a working production control system. It has three components:
(ROPE)
(Constraint)
| Element | What It Is | How to Set It |
|---|---|---|
| Drum | The constraint’s production schedule. This is the master schedule for the entire facility. | Schedule the constraint to 100% of its exploited capacity. Every other schedule is derived from this one. |
| Buffer | A time buffer of WIP upstream of the constraint. Not inventory for its own sake — insurance against upstream disruptions. | Start with buffer = 50% of the lead time from material release to constraint. Adjust based on buffer penetration data. |
| Rope | A signal from the constraint to the first operation: “I have consumed one unit — release one unit.” | Material release rate = constraint consumption rate. No more. This is what prevents WIP explosion. |
Scenario: Constraint capacity = 22 units/day. Customer demand = 20 units/day. Lead time from material release to constraint = 6 working days.
Step 1: Set the Drum.
Schedule the constraint for 20 units/day (match demand, not max capacity — the remaining 2 units/day of capacity is your exploitation buffer for recovery from disruptions).
Step 2: Set the Buffer.
Initial buffer = 50% of upstream lead time = 50% × 6 days = 3 days. Buffer size in units = 3 days × 20 units/day = 60 units of WIP in the buffer zone.
Step 3: Set the Rope.
Release material at 20 units/day. When the constraint finishes 20 units, the rope signals the release of 20 more units at the first operation. The buffer maintains ~60 units between the first operation and the constraint.
Buffer management:
- Green zone (buffer >67% full): Normal operation. No action needed.
- Yellow zone (buffer 33–67% full): Monitor. Upstream may be falling behind. Check for disruptions.
- Red zone (buffer <33% full): Expedite. The constraint is at risk of starving. Priority intervention at upstream operations.
Over time: If the buffer never enters the yellow zone, it is too large — reduce it. If it frequently enters the red zone, it is too small — increase it. The buffer finds its natural size through observation.
💡 A Buffer in Front of the Constraint Is Not Waste
A buffer in front of the constraint is not waste — it is insurance against the most expensive thing that can happen in your facility: the constraint stopping. The cost of a 60-unit buffer (inventory carrying cost) is trivial compared to the cost of constraint starvation (8+ units of lost production per hour at the constraint). Size the buffer to protect the constraint, not to minimize inventory for its own sake.
Common Failure Modes in Constraint Management
| Failure Mode | What Happens | Why It Happens | The Fix |
|---|---|---|---|
| Elevating before exploiting | Organization buys a second machine while the first has 25% untapped capacity in changeover and break losses | Capital investment feels decisive; process improvement feels slow | Demonstrate exploitation potential with data before authorizing capital |
| Improving non-constraints | Kaizen events on machines that are not the bottleneck; zero impact on facility output | Each department optimizes locally; no system-level constraint awareness | Educate leadership on system-level throughput vs. local efficiency |
| Constraint moves undetected | After successful elevation, the constraint shifts to a new operation but all improvement focus remains on the old bottleneck | Step 5 (repeat) is forgotten once the celebration ends | Monthly constraint identification review; track queue depths as leading indicator |
| Subordination rejected | Non-constraint departments refuse to reduce output; managers protect their utilization metrics | Management incentives reward local efficiency, not system throughput | Change the metrics. Reward throughput and delivery, not utilization. See Guide 14. |
⚠️ Improving a Non-Constraint Does Not Increase Output
Improving a non-constraint does not increase factory output. It creates the illusion of progress while the real constraint remains untouched. If your facility’s constraint can produce 22 units/day, no amount of improvement at any other operation will produce unit 23. The only path to unit 23 is through the constraint — either by exploiting it (Steps 2–3) or elevating it (Step 4).
🎯 The Bottom Line
The Theory of Constraints provides the decision framework that tells you where to focus. Little’s Law tells you what your lead time is. Kingman’s Equation tells you why queues form. TOC tells you where to act. Together, they form the complete physics of production flow. Next: CONWIP — the practical production control system that implements these physics by controlling what enters the system.
Interactive Demo
Add capacity anywhere in the line and watch what happens to throughput. Only one of the stations changes the answer.
Stop reading, start modeling
Model your process flow, run simulations, optimize staffing with TOC math, and test your knowledge with 107 interactive checks — all in one platform.