You Don’t Have a Speed Problem — You Have a WIP Problem
Here’s a scene I’ve witnessed dozens of times on factory floors: a plant manager staring at a whiteboard full of late orders, turning to the team, and saying, “We need to go faster.”
More overtime. More expediting. More pressure. And six weeks later? The same late orders. The same firefighting. Sometimes worse.
The problem was never speed. The problem was WIP.
Work-in-process — the inventory sitting between your first operation and your finished goods dock — is the silent killer of manufacturing lead time. And until you understand why, no amount of overtime or capital equipment will save you.
Little’s Law: The Math That Changes Everything
In the 1960s, MIT professor John Little proved a deceptively simple formula that governs every production system on earth:
Lead Time = WIP ÷ Throughput
Or equivalently:
L = W / λ
Where:
- L = average lead time (how long an item spends in your system)
- W = average work-in-process (items in the system at any point)
- λ (lambda) = average throughput (items completed per unit of time)
This isn’t a theory. It’s a law — as reliable as gravity. It holds for assembly lines, job shops, batch processes, and even product development pipelines.
Here’s what it tells us: if your throughput stays constant, every additional unit of WIP adds directly to your lead time. Not sometimes. Always. Mathematically.
Let me show you what this looks like in the real world.
The WIP Trap: Four Manufacturing Scenarios
1. Automotive Assembly: The Overloaded Line
An automotive parts supplier runs a machining-and-assembly cell. Throughput is 40 units per hour — that’s their bottleneck, and no amount of wishing changes it.
- Scenario A: 80 units of WIP on the floor → Lead Time = 80 ÷ 40 = 2 hours
- Scenario B: Manager releases 200 units “to keep everyone busy” → Lead Time = 200 ÷ 40 = 5 hours
Same throughput. Same equipment. Same people. But lead time jumped from 2 hours to 5 hours because someone flooded the floor with work. Now every job takes longer, expediting increases, and customers start screaming.
2. Electronics: The PCB Bottleneck
A PCB assembly plant runs SMT placement at 120 boards per shift. The manager launches 600 boards into the system at shift start because “we can’t afford idle time at the front end.”
- Lead Time = 600 ÷ 120 = 5 shifts (nearly a full week)
Cut the release to 240 boards with controlled pulls:
- Lead Time = 240 ÷ 120 = 2 shifts
The customer gets boards in 2 days instead of 5. Quality improves too — defects get caught 3 shifts sooner. The rework pile shrinks. WIP carrying costs drop.
3. Food Processing: The Perishable Problem
A frozen food plant packs 500 cases per hour. During a promotional spike, planning pushes 4,000 cases into the system:
- Lead Time = 4,000 ÷ 500 = 8 hours
For a perishable product, those extra hours in a non-refrigerated staging area create real quality risk. Pull it back to 1,500 cases:
- Lead Time = 1,500 ÷ 500 = 3 hours
Less spoilage. Less rework. Fewer customer complaints. And they still shipped the same volume — just in smaller, faster batches.
4. Pharma: The Compliance Cascade
A pharmaceutical packaging line runs 200 units per hour through serialization. When 1,600 units are staged ahead of the line:
- Lead Time = 1,600 ÷ 200 = 8 hours
In pharma, longer lead times mean more time for environmental excursions, more lot tracking complexity, and more deviation risk. Reducing WIP to 600 units:
- Lead Time = 600 ÷ 200 = 3 hours
Same output. Less compliance risk. Simpler batch records.
Why Adding WIP Never Increases Output
This is the trap managers fall into every single day. The logic seems sound: “If I put more work into the system, more work will come out.”
But Little’s Law says otherwise. Throughput is governed by your bottleneck — the slowest operation in your process. Releasing more work into the system doesn’t speed up the bottleneck. It just stacks inventory in front of it.
Here’s the math, plain and simple:
| WIP (units) | Throughput (units/hr) | Lead Time |
|---|---|---|
| 100 | 50 | 2 hours |
| 200 | 50 | 4 hours |
| 400 | 50 | 8 hours |
| 800 | 50 | 16 hours |
Throughput doesn’t change. It can’t change just because you released more work. All you did was multiply your lead time — and every problem that comes with it: congestion, confusion, quality defects, expediting, and overtime to dig out of the mess you created by “staying busy.”
I’ve seen this play out on factory floors across automotive, aerospace, medical devices, and consumer goods. The pattern is always the same.
High-WIP vs. Low-WIP: A Side-by-Side Comparison
| Metric | High-WIP Plant | Low-WIP Plant |
|---|---|---|
| Lead Time | 3–5× quoted | At or below quoted |
| On-Time Delivery | 60–75% | 92–98% |
| Expediting Cost | 8–15% of labor budget | < 2% |
| Quality (first-pass yield) | 82–88% | 94–97% |
| Floor Space Utilization | Cluttered, aisles blocked | Clean, visual, organized |
| Overtime | Chronic (15–25%) | Occasional (< 5%) |
| Inventory Carrying Cost | High (tied-up capital) | Low (fast turns) |
| Employee Morale | Firefighting mode | Predictable, focused |
The low-WIP plant doesn’t have better equipment or faster workers. It has discipline — the discipline to limit what enters the system to what the system can actually process.
How to Reduce WIP: Practical Steps
Knowing the problem is one thing. Fixing it requires specific tools and the courage to fight the “stay busy” culture. Here’s what works:
1. Set Explicit WIP Limits (Kanban)
Use kanban cards or digital signals to cap the number of jobs at each workstation. When a downstream station finishes a unit, it sends a signal upstream: “I’m ready for the next one.” No signal, no release.
This is pull-based production, and it’s the single most powerful WIP-control mechanism in lean manufacturing.
2. CONWIP (Constant Work-in-Process)
For job shops and high-mix environments where station-level kanban is impractical, CONWIP sets a system-wide WIP cap. A new job enters the system only when a finished job exits. Simple, robust, and effective.
3. Drum-Buffer-Rope (Theory of Constraints)
Eliyahu Goldratt’s method ties the release rate to the bottleneck’s capacity (the “drum”). A time buffer protects the bottleneck from starvation. A “rope” pulls new work into the system at exactly the rate the constraint can process it.
This is especially powerful in plants with a clear, persistent bottleneck — heat treat, test, or a single specialized machine.
4. Reduce Batch Sizes
Large batches are a WIP multiplier. If your bottleneck runs 50 units/hour and you release batches of 500, you’ve just guaranteed 10 hours of queue time before the last unit even starts. Cut to batches of 100, and the math transforms.
5. Fix the Bottleneck
After you control WIP, then invest in the bottleneck. Improve changeover times (SMED). Add capacity at the constraint. Eliminate downtime. Every minute gained at the bottleneck increases λ in Little’s Law — and that’s the only way to sustainably reduce lead time while maintaining volume.
But What If You Have Multiple Servers? Enter M/M/c
Everything above assumes a single process or bottleneck. But real operations often have parallel servers — multiple assembly stations, multiple technicians, multiple machines doing the same job. This is where the M/M/c queueing model comes in.
Little’s Law still holds (it always does), but the dynamics change fundamentally when you add servers.
The M/M/c Model
The notation tells you the setup:
- M = Markovian (random) arrivals — Poisson process
- M = Markovian (random) service times — exponential
- c = number of identical parallel servers
The key insight: adding a server doesn’t just add capacity linearly — it reduces wait times exponentially. This is because of the Erlang C formula, which calculates the probability that an arriving job has to wait.
Manufacturing Example: Assembly Stations
A plant has 20 assemblies arriving per hour, each station processes 8 per hour:
| Servers | Utilization | P(Wait) | Avg Queue | Avg Wait Time |
|---|---|---|---|---|
| 3 | 83% | 65% | 2.6 jobs | 7.8 min |
| 4 | 63% | 22% | 0.4 jobs | 1.2 min |
| 5 | 50% | 7% | 0.07 jobs | 0.2 min |
Look at the jump from 3 to 4 servers: utilization drops from 83% to 63%, but wait time drops by 85% — from nearly 8 minutes to just over 1 minute. That’s the nonlinear magic of multi-server queueing.
The Counterintuitive Truth About Utilization
Most managers want 95%+ utilization — “keep everyone busy.” But M/M/c theory proves that high utilization creates exponentially longer queues. At 90% utilization, queue times explode. At 95%, they’re catastrophic.
The sweet spot for most manufacturing operations is 70–85% utilization. Below that, you’re overstaffed. Above that, wait times spiral.
This is why adding one extra server (or machine, or technician) at the right moment can slash lead times by 50%+ — not because you added 33% capacity, but because you moved off the steep part of the waiting curve.
When to Use Which Model
| Situation | Use This |
|---|---|
| Single bottleneck, single flow | Little’s Law (L = λW) |
| Parallel machines/people, same job | M/M/c with Erlang C |
| Right-sizing a team or machine pool | M/M/c — find the minimum c for your target wait time |
| General WIP analysis, any system | Little’s Law — always valid |
Both tools are available in our Operations Calculator — use the Little’s Law tab for flow analysis and the Multi-Server Queue (M/M/c) tab for staffing and capacity decisions.
The Lean Six Sigma Connection
If you’ve studied Lean Six Sigma, everything here connects. WIP reduction is where Lean (flow, pull, waste elimination) meets Six Sigma (variation reduction, process capability).
High WIP is a symptom of variation — in demand, processing time, quality, and scheduling. Six Sigma tools like control charts, capability studies, and DMAIC projects attack the causes of WIP accumulation. Lean tools like kanban, value stream mapping, and standard work create the system that holds the gains.
As a Lean Six Sigma practitioner, I’ve found that the fastest ROI in almost any manufacturing operation comes from a single project: map the value stream, identify the bottleneck, set a WIP limit, and enforce it. Lead times drop 30–60% within weeks. Not months. Weeks.
Stop Going Faster. Start Controlling WIP.
The next time someone in your plant says “We need to speed up,” ask them one question:
“What’s our current WIP?”
If they don’t know the answer — and most don’t — that’s your first problem. You can’t manage what you can’t see.
Start here:
- Count your WIP. Right now. Physically.
- Calculate your current lead time using Little’s Law.
- Set a WIP limit 20–30% below current levels.
- Enforce it for 30 days.
- Measure lead time, on-time delivery, and quality.
The results will speak for themselves.
Want to run the numbers for your operation? Use our Operations Calculator to model WIP scenarios with Little’s Law and right-size your team with the M/M/c multi-server queue analyzer.
JJ Andrade is a Business Performance Mentor and Lean Six Sigma expert who helps manufacturers eliminate waste, reduce lead times, and build systems that scale. Get in touch to discuss your WIP challenge.