A fabrication shop owner in Pune was convinced his welding section was the problem. Every delivery was late. Overtime was through the roof. He had six welders working double shifts and was about to hire two more. Then someone looked at the shop floor carefully — not at the welding bays, but at what was happening before and after them.
The welding bays were full, yes. But the jobs sitting in front of them were waiting for powder coating, not welding. The real bottleneck was the powder coating booth — a single oven that could process only 12 batches per day. Welding was finishing faster than powder coating could handle. The WIP pile-up in front of welding was actually overflow from the powder coating queue, pushed backward because there was no floor space left near the oven.
He almost hired two welders he did not need. The fix was increasing powder coating throughput — which he did by adding a second shift for the oven and pre-staging batches. Cost: ₹40,000 per month in overtime and one helper's salary. The alternative — two welders at ₹50,000 each plus equipment — would have cost ₹1.5 lakh per month and solved nothing.
This is what bottleneck analysis does. It stops you from working hard on the wrong machine.
The Theory of Constraints — without the textbook
The Theory of Constraints (TOC) was developed by Eliyahu Goldratt, most famously explained in his book about a factory manager who figures out that his entire factory's output is limited by one machine. The core insight is simple and powerful:
Every factory has one constraint — one resource that limits the output of the entire system. Improving anything other than the constraint does not improve the factory's output.
This sounds obvious when stated plainly. In practice, factories violate it every day. They buy faster machines that are not the bottleneck. They add overtime on workstations that are already ahead of the constraint. They invest in automation for operations that have spare capacity. The money is spent. The output does not change. The delivery dates do not improve.
Why Indian SMEs need this thinking
Indian MSMEs typically operate with tight capital and limited floor space. You cannot afford to invest in the wrong place. A CNC machine costs ₹15-40 lakh. Hiring and training a skilled operator takes 3-6 months and ₹3-5 lakh per year in salary. Floor space in industrial areas of Pune, Faridabad, or Bengaluru costs ₹15-30 per square foot per month.
Every investment that does not increase throughput at the constraint is wasted. TOC gives you a framework to make sure your next ₹10 lakh goes to the right place.
How to identify the bottleneck
The bottleneck is not the busiest machine. It is not the machine that breaks down the most. It is not the machine the operators complain about the most. The bottleneck is the machine where work-in-progress accumulates.
The WIP test
Walk your shop floor. Look at where semi-finished goods are piling up. Not raw materials (that is a purchasing problem) and not finished goods (that is a sales problem). Look for WIP — parts that have completed some operations but are waiting for the next one.
| What you see | What it means |
|---|---|
| Large WIP pile before a machine/workstation | This machine is likely the bottleneck — work arrives faster than it can process |
| Empty queue before a machine | This machine has more capacity than it needs — it is NOT the bottleneck |
| WIP spread evenly across the floor | Either the flow is well-balanced (rare) or you have multiple constraints (review further) |
| Finished goods piling up after a machine | The downstream operation is the bottleneck, not this one |
The queue time test
If you have basic production tracking in place, look at the data. For each operation in your routing, calculate the average queue time — the time a job spends waiting to be processed after the previous operation is complete.
The operation with the longest average queue time is sitting in front of the bottleneck. If jobs wait 3 hours for CNC turning but 2 days for powder coating, powder coating is your constraint, regardless of how busy the CNC lathe looks.
The overtime test
Which machine or workstation consistently requires overtime to keep up? If welding works overtime 4 days out of 5 while every other station manages within normal shifts, welding is probably the constraint.
But be careful — overtime can mask the real bottleneck. If the powder coating booth runs only one shift (because the owner thinks it does not need two shifts), it might appear to keep up during the day while jobs silently pile up overnight. Check queue lengths, not just overtime hours.
The "what if" test
Ask your production manager: "If I could magically add 50% more capacity to one machine, which one would improve our delivery performance the most?" The answer they give you — without analysis, from pure shop-floor intuition — is usually correct. Production managers know the bottleneck. They just do not always call it that.
The five focusing steps
TOC provides a systematic five-step process for improving throughput. These steps are meant to be followed in sequence — and the sequence matters.
Step 1: Identify the constraint
Use the methods above. Find the one resource that limits your factory's output. In a fabrication shop, common constraints are:
- Powder coating / painting — limited by oven cycles, curing time, and floor space
- CNC machining — limited by available machine hours, especially for precision work
- Welding — limited by skilled welder availability, especially for TIG/specialised welding
- Assembly — limited by floor space, fixture availability, or inspection bottlenecks
- Inspection/quality — limited by CMM availability, inspector bandwidth
Sometimes the constraint is not a machine at all. It can be:
- A specific skill — you have one person who can read complex GD&T drawings
- A shared resource — the overhead crane, the paint booth, the test bench
- Information — waiting for customer approvals, drawing revisions, or material test certificates
- Material — a specific raw material with long lead times or unreliable suppliers
Step 2: Exploit the constraint
Before spending any money, squeeze every minute of capacity out of the bottleneck. This is the cheapest step and often the most impactful.
Exploitation tactics for a machine bottleneck:
Eliminate idle time on the constraint. The constraint should never wait for anything — not for material, not for the operator, not for the crane, not for instructions. Stage the next job's material, fixtures, and programs before the current job finishes. If the constraint operator needs to go for lunch, have a relief operator take over. Zero idle minutes.
Stagger breaks. If the powder coating oven is the bottleneck, the oven operator does not take lunch when everyone else does. Someone covers the oven during lunch. A bottleneck minute is a factory minute — you cannot get it back.
Reduce changeover time on the constraint. Apply basic SMED (Single-Minute Exchange of Die) principles. Separate external setup (what can be done while the machine is running) from internal setup (what must be done with the machine stopped). In a typical Indian job shop, 30-50% of changeover activities can be moved to external setup with zero investment.
Ensure quality before the constraint. Every defective part that reaches the constraint wastes constraint time. Implement an inspection step before the constraint — check dimensions, verify material, confirm the drawing. If a part is wrong, reject it before it consumes bottleneck capacity, not after.
Prioritise high-value jobs on the constraint. When the constraint has a queue of jobs waiting, process the highest-revenue-per-hour job first. If Job A generates ₹5,000 per hour of constraint time and Job B generates ₹2,000 per hour, process Job A first. The factory makes more money with the same constraint capacity.
Step 3: Subordinate everything else to the constraint
This is the step that most factories skip, and it is the most important one.
Subordination means that every other resource in the factory adjusts its pace and priorities to support the constraint. Non-constraint resources should not be optimised independently — they should be optimised to keep the constraint fed and productive.
What subordination looks like in practice:
Upstream machines (before the constraint) produce only what the constraint needs next. Do not let the cutting machine run ahead and build a 2-week buffer of cut parts. Cut what the constraint will process in the next 1-2 days. Excess WIP clutters the floor, ties up capital, and creates confusion.
Downstream machines (after the constraint) process whatever the constraint produces, immediately. If the constraint finishes a batch at 3 PM, the next station should start processing it by 3:30 PM, not the next morning. The constraint's output is the factory's most valuable flow — do not let it sit.
Maintenance prioritises the constraint. If two machines need repair and one is the constraint, the constraint gets fixed first. Always. The non-constraint machine can wait — its downtime does not reduce factory output.
Scheduling prioritises the constraint. The daily schedule should be built around the constraint's capacity. Decide what the constraint will work on, then schedule everything else to feed it and process its output.
Step 4: Elevate the constraint
If exploitation and subordination are not enough — if the constraint is still limiting your output even after you have squeezed every minute out of it — then invest in increasing its capacity.
Elevation tactics:
- Add a shift. The cheapest way to increase constraint capacity by 50-100%. If the powder coating oven runs one shift, adding a second shift doubles its capacity for the cost of one operator's overtime salary.
- Outsource the constraint operation. Send some powder coating to an external vendor. The per-unit cost is higher, but you free up internal capacity for jobs that cannot be outsourced.
- Buy additional equipment. A second powder coating oven. An additional CNC lathe. This is the most expensive option and should be the last resort, not the first reaction.
- Invest in efficiency improvements. Better tooling, faster fixtures, automation for loading/unloading. These increase the effective capacity of the existing machine.
Cost comparison for elevating a powder coating bottleneck:
| Option | One-time cost | Monthly cost | Capacity increase |
|---|---|---|---|
| Add second shift (overtime + helper) | ₹0 | ₹40,000-60,000 | +50-80% |
| Outsource 30% of volume | ₹0 | ₹50,000-80,000 (variable) | +30% |
| Improve curing cycle (better oven control) | ₹1-2 lakh | Negligible | +10-15% |
| Buy second oven | ₹8-15 lakh | Maintenance costs | +80-100% |
In the Pune fabrication shop example, adding a second shift cost ₹45,000 per month and freed up enough capacity to take on ₹4-5 lakh per month of additional work. The ROI was immediate.
Step 5: Go back to Step 1
After you elevate the constraint, it may no longer be the constraint. The bottleneck shifts. The powder coating booth now has enough capacity, but CNC machining has become the new constraint because the factory is processing more jobs overall.
This is the cycle of continuous improvement. Find the constraint. Exploit it. Subordinate everything to it. Elevate it if needed. Find the new constraint. Repeat.
Warning: Do not automate this cycle. Each time the constraint shifts, the factory dynamics change. The exploitation and subordination tactics for a CNC machining constraint are completely different from those for a powder coating constraint. Re-assess each time.
The Pune fabrication shop: a detailed case study
Let's walk through the example from the introduction in more detail, because it illustrates several common mistakes.
The situation
A fabrication shop in Pune, specialising in steel structures and enclosures for the electrical panel industry. The factory has:
- 2 CNC plasma cutting tables
- 4 press brakes (bending)
- 6 MIG welding bays
- 1 powder coating line (1 oven, 1 spray booth)
- 1 inspection and packing area
- 35 workers across two shifts
The symptoms:
- On-time delivery rate: 55%
- Average delay on late orders: 8-12 days
- Overtime: 4-5 days per week, mainly on welding
- Floor WIP: high, cluttered, hard to move material around
- Monthly throughput: ₹45-50 lakh (felt capped — could not grow beyond this)
The incorrect diagnosis
The owner saw welding working overtime every week and concluded: "Welding is the bottleneck. I need more welders." He was preparing to hire two additional welders at ₹25,000 each per month and add a 7th welding bay (₹2 lakh for equipment and setup).
The correct diagnosis
A careful analysis of the shop floor revealed:
| Workstation | Daily capacity (hours) | Average daily load (hours) | Utilisation | Average queue (days) |
|---|---|---|---|---|
| CNC Plasma | 32 (2 tables × 16h) | 22 | 69% | 0.3 days |
| Press Brake | 64 (4 machines × 16h) | 38 | 59% | 0.2 days |
| Welding | 96 (6 bays × 16h) | 78 | 81% | 0.8 days |
| Powder Coating | 20 (1 line, limited by oven) | 22 | 110% (overloaded) | 3.2 days |
| Inspection/Packing | 16 | 8 | 50% | 0.1 days |
The powder coating line was running at 110% utilisation — meaning it could not keep up even with overtime. The average queue before powder coating was 3.2 days. Every job that passed through welding sat for more than 3 days waiting for the oven.
Welding appeared busy because jobs were backing up from powder coating into the welding area. Welders were working overtime not because welding was the constraint, but because jobs were being pushed to welding faster than the powder coating could clear them downstream. The welding area was full of completed weldments waiting for coating — and new weldments waiting for floor space to start.
The fix
Exploitation (Week 1-2):
- Staggered the powder coating operator's lunch break (30 minutes of extra oven time per day)
- Pre-staged parts for coating the night before (reduced oven idle time between batches by 15 minutes per batch, saving ~2 hours per day)
- Implemented quality check before coating (reduced rework after coating from 8% to 2% — each rework consumed one full oven cycle)
These three changes increased effective powder coating capacity from 20 hours to approximately 25 hours per day — for zero capital investment.
Subordination (Week 2-3):
- Welding was instructed to prioritise jobs that were waiting longest for powder coating, not jobs that were newest. This reduced the average powder coating queue from 3.2 days to 1.5 days.
- Plasma cutting was paced to not outrun the welding-to-coating flow. Previously, plasma ran at full speed and created floor clutter. Now it was scheduled to feed welding at the rate powder coating could absorb.
Elevation (Week 3-4):
- Added a second shift for the powder coating oven: one helper at ₹18,000/month + overtime for the coating operator at ₹22,000/month additional = ₹40,000/month.
- This increased daily coating capacity from 25 hours (exploited) to approximately 38 hours.
The results (after 90 days)
| Metric | Before | After | Change |
|---|---|---|---|
| On-time delivery | 55% | 82% | +27 percentage points |
| Average delay (late orders) | 8-12 days | 2-3 days | -70% |
| Weekly overtime (welding) | 4-5 days | 1-2 days | -60% |
| Monthly throughput | ₹45-50 lakh | ₹62-65 lakh | +35% |
| Floor WIP (visual estimate) | Very high | Moderate | Significant reduction |
| Additional monthly cost | — | ₹40,000 (2nd shift coating) | — |
| Additional monthly revenue | — | ₹15-17 lakh | 37x return on ₹40,000 |
The ₹40,000 per month investment in second-shift powder coating generated ₹15-17 lakh in additional monthly throughput. Meanwhile, the proposed ₹1.5 lakh/month investment in additional welders would have generated approximately zero additional throughput — because welding was never the constraint.
The diagnostic checklist
Use this checklist to find your factory's constraint. Walk the floor with a clipboard and answer these questions:
Physical observation (30 minutes)
- Where is WIP visually accumulating? Mark the workstations.
- Which machines have empty queues (waiting for work)?
- Which machines have overflowing queues (work waiting for the machine)?
- Is any material stored in aisles or unusual places because there is no room near the designated area?
- Which machine or area consistently runs overtime?
Data check (if production tracking is available)
- What is the average queue time before each operation? Which is highest?
- What is the utilisation of each key machine/workstation? Which exceeds 85%?
- Which machine has the most unplanned downtime?
- Which operation has the highest rejection/rework rate?
Interview check (talk to people)
- Ask the production manager: "If you could add capacity to one machine, which one would help the most?"
- Ask operators: "Which machine does everyone wait for?"
- Ask the dispatch/sales team: "When deliveries are late, which operation is usually the last to finish?"
Constraint identification
- The workstation with the longest queue time is likely the constraint.
- If multiple workstations have long queues, identify which one feeds the others — the upstream constraint creates the downstream pileup.
- Verify: if you could magically add 50% capacity to this one workstation, would factory throughput increase? If yes, you have found the constraint.
Common false constraints
Watch out for these red herrings:
The machine that breaks down a lot. A machine with frequent breakdowns is frustrating but is only a constraint if it also has the longest queue. A press brake that breaks down twice a week but has only 4 hours of daily load is not a constraint — it just needs maintenance.
The most expensive machine. The ₹35 lakh CNC 5-axis is not automatically the bottleneck. If it runs 60% utilised while the ₹3 lakh manual lathe has a 4-day queue, the lathe is your constraint.
The machine the owner is most proud of. Emotional attachment to equipment distorts analysis. Analyse the flow, not the feelings.
The newest hire. A slow operator on a non-bottleneck machine does not limit factory output. A slow operator on the bottleneck machine limits everything. Focus training investments on constraint operators.
Throughput accounting — thinking in constraint minutes
Once you have identified the constraint, every decision in the factory should be evaluated in terms of constraint minutes.
Throughput per constraint minute
For each product or job, calculate:
Throughput per constraint minute = (Selling price - Material cost) ÷ Minutes on constraint
Example with three jobs competing for the CNC lathe (the constraint):
| Job | Selling price | Material cost | Throughput | CNC time (min) | Throughput per CNC minute |
|---|---|---|---|---|---|
| Job A: Shaft batch | ₹1,80,000 | ₹65,000 | ₹1,15,000 | 480 min | ₹240/min |
| Job B: Housing batch | ₹2,40,000 | ₹1,10,000 | ₹1,30,000 | 720 min | ₹181/min |
| Job C: Bracket batch | ₹90,000 | ₹30,000 | ₹60,000 | 180 min | ₹333/min |
Traditional thinking: Job B is the most profitable (highest total throughput of ₹1,30,000). Prioritise Job B.
Constraint thinking: Job C generates ₹333 per minute of constraint time. Prioritise Job C, then Job A, then Job B. The factory makes more money per day by maximising throughput per constraint minute.
This logic applies to quoting as well. When a customer requests a quote and the constraint is fully loaded, price the job based on the throughput it must generate per constraint minute to be worth accepting. If the constraint is idle, even a low-margin job is better than an empty machine.
Applying this to your factory — today
You do not need a consultant. You do not need software. You need 30 minutes and a willingness to look at your factory with fresh eyes.
- Walk the floor. Look for WIP pileups. Where is semi-finished inventory accumulating?
- Identify the constraint. Use the checklist above. Talk to your production manager.
- Exploit it. Find 2-3 ways to squeeze more capacity from the constraint without spending money. Stagger breaks, pre-stage material, improve quality upstream.
- Subordinate to it. Tell upstream operations to slow down to the constraint's pace. Tell downstream operations to prioritise the constraint's output.
- Measure. Track the constraint's utilisation daily. Track factory throughput weekly. See if the numbers move.
The factory that finds and fixes its real constraint does not just improve one metric. It improves everything downstream: throughput goes up, WIP goes down, lead times shrink, delivery performance improves, and overtime drops. All from focusing on one machine instead of fighting fires on all of them.
QuoteERP helps you connect bottleneck data to your quoting — so you price jobs based on real constraint capacity and make decisions that maximise throughput, not just revenue. Find out how →