A factory can have hundreds of problems, but they don't all restrict output.

To diagnose why a factory isn't performing, I start by understanding the type of production system it is, what it needs to deliver and how work really flows through it. Then I look for the bottleneck controlling output, why that constraint exists, and the behaviours and management beliefs that allow the problem to continue.

The objective isn't to find everything that could be improved.

It's to find what is restricting the performance of the factory and what should be fixed first.

That distinction matters.

A factory can be extremely busy and still produce too little.

You can improve the efficiency of an individual machine without increasing factory output.

You can recruit more people without increasing capacity.

And you can implement Lean, automation or new software without fixing the problem that is actually preventing orders from getting out of the door.

That's why I diagnose before I improve.

What should you look at when diagnosing a manufacturing problem?

I use nine broad stages:

Understand the operating model – what type of factory are you actually running? There are 5 types.
Understand demand and the required production rate – what does the factory actually need to make, to generate the revenue?
Follow the whole Manufacturing Process Cycle – from customer enquiry through to dispatch.
Make the flow of work visible – understand what is actually happening rather than what the system says or people think should happen.
Find the true bottleneck – what is controlling the rate of output?
Understand why the bottleneck is constrained – is the bottleneck really short of capacity or is capacity being lost?
Understand the behaviours and beliefs behind the system – People have seen and put up with the symptoms, why are people still managing the factory this way?
Identify the underlying failure mechanisms – connect individual symptoms to their common causes.
Decide what to fix first – prioritise the changes that will have the greatest effect on factory performance.

The order matters.

If you decide on the solution before understanding the problem, you're guessing.

1. Understand the Factory You're Actually Running

Before judging whether a factory is performing well, I want to understand what kind of production system it is.

Manufacturing businesses aren't all variations of the same factory.

I broadly group them into five manufacturing operating models:

Single-Stage

Raw material is loaded onto a single machine, it gets processed and might be inspected and packed. Think injection moulding or cnc milling, the order, the material doesn't go to more than 1 machine.

Capacity may appear straightforward, but numbers of machines, product mix, batch sizes, changeovers and available production time can dramatically affect actual output.

Sequential

Products pass through a broadly repeated sequence of manufacturing operations.

Balance between processes, variation, takt and interruptions become important because one process affects what happens next.

Parallel / Modular

Different products, assemblies or modules are produced through parallel routes.

Individual departments can look highly productive while the overall product waits for the one module that isn't ready.

Convergent / Hybrid

Multiple materials, manufactured components and sub-assemblies converge through several levels into a finished product.

This is common in machinery and complex assembly.

Planning, design, purchasing, kitting, sequencing and the timing of sub-assemblies can become as important as assembly capacity itself.

Starting more machines doesn't necessarily mean finishing more machines.

Continuous Flow

Material moves through an interconnected production process.

Availability, speed losses, interruptions, yield and the relationship between processes can determine the performance of the whole system.

These five models aren't simply labels.

They change what I expect to see when I walk into the factory.

I then consider how frequently orders occur, not just the size of the order and whether the business is primarily Make to Stock, Make to Order, Assemble to Order or Engineer to Order.

A plastics manufacturer producing thousands of relatively similar products needs to be planned and controlled differently from another plastics production facility doing short runs of 1-2 days only.

A manufacturing business building six complex engineered machines each month is vastly different from a factory filling 2,000 bottles of detergent an hour.

They all have customers, orders to deliver on-time, they employ staff, have to make to cover all the costs and deliver a profit.

Before you improve a production system, you need to understand what production system you actually need.

2. Understand What the Factory Needs to Achieve

“Production is really busy” tells me very little.

I want to understand the required rate.

If the business needs to manufacture 60 machines a year, what does that mean per month and per week?

If demand is increasing from 750,000 units to one million units per month, where does the additional output need to come from?

If product mix changes, what happens to the requirement?

How much production time is genuinely available?

What capacity is required at each important stage?

This is where averages can become misleading.

A factory may have enough total labour hours or machine hours on paper while still being unable to satisfy demand.

Those hours need to exist in the right place, at the right time and in the right sequence.

Capacity therefore can't be diagnosed simply by adding together everyone's available hours.

You have to understand how capacity behaves through the production system.

3. Follow the Whole Manufacturing Process Cycle

Manufacturing problems don't necessarily start on the shop floor.

That's why I follow the order through the whole Manufacturing Process Cycle:

Sales enquiry / order → Planning → Purchasing → Scheduling → Work order & kitting → Production / assembly / finishing → Test / inspection / commissioning → Pack / dispatch

The exact stages and the impact they have, vary from business to business.

The principle doesn't.

I want to understand how customer demand becomes a finished product.

That means following information as well as material.

When is the customer delivery date agreed?
Who determines how the product will be made?
When does design happen?
When are materials purchased?
When is capacity checked?
Who decides what should be worked on next?
When is work released?
Are kits complete when production receives them?
When is testing scheduled?
Does dispatch form part of the production plan?
And where does work wait?

This matters because a manufacturing bottleneck isn't necessarily a machine.

It could be:

Design. A particular engineer. An approval. Procurement. Kitting. A scarce skill. A production bay. A curing process. Inspection. Testing. Commissioning. Dispatch.

Sometimes it is simply a decision that only one person can make.

The constraint, the bottleneck can exist anywhere in the system that converts customer demand into delivered product.

4. Make the Factory Visible

Most factories aren't short of data.

ERP, MRP, CRM systems contain orders, routings, material transactions, stock, purchasing information, labour bookings and dates.

Spreadsheets contain another version of the plan.

Production boards contain another.

And managers frequently carry another version in their heads.

The problem is often that nobody can see the production system as a whole.

So I try to make it visible.

In a complex machine-building environment, for example, I might model:

when each machine was started;
when sub-assemblies were started;
when material was issued;
when assemblies were completed;
how many machines were being built simultaneously;
where work waited;
which resources different machines were competing for;
and what prevented each machine moving to its next stage.

Thousands of individual ERP, MRP, Spreadsheet transactions can then reveal a pattern.

What looked like ten machines progressing through production may actually be ten partially completed machines competing for the same limited resources, at the same time.

That leads to a completely different diagnosis.

5. Find the True Bottleneck

What is a manufacturing bottleneck?
A manufacturing bottleneck is the resource, process or decision that controls the rate at which the production system can satisfy customer demand. It isn't necessarily a machine. It can exist anywhere in the Manufacturing Process Cycle.

This is one of the most important questions I ask:

What is actually controlling the rate at which this factory can produce what customers need?

Think about an egg timer.

There can be plenty of sand waiting above the neck, but the amount coming out is controlled by the narrowest point.

A factory behaves in much the same way.

If one process can produce 100 units per hour but the next can only process 60, increasing the first process to 110 doesn't increase factory output.

It creates more work-in-progress.

Factory performance comes from flow, not busyness.

The bottleneck isn't always obvious.

I look for evidence such as persistent queues, downstream processes waiting for work, concentrated overtime, repeated expediting, excessive WIP, scarce skills, limited equipment, incomplete kits and work continually being reprioritised.

But identifying the apparent bottleneck isn't enough.

The more important question is:

Why is it the bottleneck?

6. Understand Why the Bottleneck Is Constrained

Finding a busy process doesn't prove that the factory needs more capacity.

Suppose welding appears to be the bottleneck.

The obvious answer might be:

“We need another welder.”

Maybe.

But what are the existing welders actually doing?

Are they welding?

Or are they measuring, locating, moving material, searching for components, setting up, waiting for information and correcting earlier errors?

Can a manufacturing bottleneck be something other than a machine?

The same applies outside production.

Testing may appear short of capacity because machines arrive incomplete.

Design may appear overloaded because engineers are continually interrupted by production queries.

Purchasing may appear to be causing shortages because work is being released much earlier than the factory can consume it.

A bottleneck can represent a genuine shortage of capacity.

But it can also represent existing capacity being consumed badly.

Those require very different solutions.

7. Understand Why People Are Managing the Factory This Way

This is an important part of diagnosis that is easily overlooked.

Factories are run by people. The systems you see are partly the result of their experience, training, beliefs and understanding of manufacturing.

Why do manufacturing problems persist?

Manufacturing problems often persist because people continue managing according to the experience, training, measures and mental models they already have. If nobody has shown them a different way to manage flow, capacity or production, the existing system can continue even when everyone can see its symptoms.

So I don't only ask:

“Why does this happen?”

I also ask:

“Why do people do it this way?”

And:

“Why haven't they changed it? They've seen the problems, they see them everyday”

A Production Manager may believe keeping every machine busy is the most efficient way to run the factory.

A supervisor may believe good management means constantly monitoring people.

A planner may schedule every machine independently because that's how they were trained.

Someone may recognise that the current approach isn't working but have no experience of a different production system.

Others may be extremely capable technically or administratively but less confident leading people, challenging established practice or making decisions when the answer isn't obvious.

Experience matters too.

What manufacturing environments have they worked in before?
What formal training have they received?
Who has mentored them?
Have they ever been shown how to identify a constraint, manage WIP, schedule a convergent assembly operation or control production through work release?

People often solve the problem they understand, not necessarily the problem the factory has.

That's important because some manufacturing problems are genuinely difficult.

Finding the true bottleneck in a complex factory can be difficult.

Recognising that a plastics factory shouldn't primarily plan individual machines but should plan the sequence and frequency of its changeovers requires a different mental model.

Recognising that a multi-stage factory should plan from its final production operation and release orders at the rate that constraint can work isn't necessarily intuitive either.

If nobody in the business has been taught those approaches or experienced them elsewhere, why would we expect them to invent them?

This is why I look at the management system and the production system together.

Changing a process without changing the thinking that created will mean the old process gradually returns and that means the problems will too.

8. Identify the Underlying Failure Mechanism

Individual symptoms, observations rarely exist in isolation.

A late material delivery, incomplete kit, changing production schedule and machine waiting in assembly may look like four different problems.

They may actually be symptoms of the same underlying failure.

Across factories I repeatedly find mechanisms such as:

Demand isn't translated into a realistic production requirement.
Work is released before it is ready.
The schedule doesn't control priorities and flow.
The factory doesn't see and act far enough ahead.
Local activity is prioritised over overall factory flow.
The real constraint isn't understood or managed.
Managers are compensating for weaknesses in the operating system.
The production system hasn't been consciously designed for the factory being operated.
The manufacturing process itself isn't sufficiently defined, stable or repeatable.

This distinction matters.

If you attack every symptom independently, you can easily create dozens of improvement projects. Projects that add work to the current workload for everyone.

If you identify the mechanism creating several symptoms, one intervention may solve several problems at once.

9. Decide What to Fix First

A useful factory diagnosis shouldn't end with 47 recommendations.

Most management teams already know dozens of things that could be better.

The difficult question is:

Which changes will make the biggest difference to factory performance?

I use four considerations:

Impact

Will solving it materially improve throughput, delivery, lead time, capacity, cost or another important business result? Not just make a single machine faster, a team more efficient.

Cause

Are we fixing a real underlying cause or just a single symptom?

Speed

How quickly can the business implement the change and see a result? Can the manager do this as part of their day to day job? What can we stop doing while we do this?

Unlock

Will solving this problem make other improvements possible? Will it create time that can be spent on other improvements.

I then broadly separate recommendations into:

Fix First – the small number of issues currently restricting factory performance.

Fix Next – changes that become valuable once the immediate constraints are addressed.

Fix Later – worthwhile improvements that shouldn't distract management from the bigger problem.

The result shouldn't be an improvement wish list.

It should be an order of attack.

Why I Diagnose Before Recommending Lean, Automation or More Capacity

Only after understanding the production system do I start deciding which tools or interventions it needs.

The answer might be Lean. It might be

5S, SMED, standard work, Kanban or pull.
better production scheduling.
changing batch sizes or production sequence.
controlling work release.
a different factory layout.
management training or coaching.
changing ERP parameters.
automation.

And sometimes the factory genuinely does need another machine or more people.

But the solution comes after the diagnosis.

I've seen businesses trying to solve flow problems with additional capacity and capacity problems with better planning.

Have you ever put a 4 hour Saturday morning shift on, an extra 10% of a 40 hour week, only to find you only got 5% more work?

I've seen individual processes improved without materially increasing the amount of finished product leaving the factory. That improvement idea that claimed to have increased a machine efficiency by 20% but no extra orders left the factory.

Improvement should improve the business result, not just the process.

What Should a Factory Diagnosis Tell You?

At the end of a good manufacturing diagnosis, the leadership team should be able to answer:

What type of production system are we actually running?
What does customer demand require it to achieve?
How does work really flow from enquiry to dispatch?
Where is the true bottleneck?
Why is it constrained?
Which behaviours, beliefs or management systems are reinforcing the problem?
What are the underlying causes rather than the visible symptoms?
What should we fix first?

And sometimes just as importantly:

What don't we need to spend money on yet?

The answer to a manufacturing problem isn't automatically another person, another machine, another piece of software or another improvement programme.

Sometimes the capacity already exists.

The factory simply isn't converting enough of its available time into throughput.

Diagnose the Factory You're Actually Running

Every factory is different.

That's precisely why I don't believe in walking into one with the solution already chosen.

But different factories can still be diagnosed systematically.

Understand the operating model.
Understand demand.
Follow the whole Manufacturing Process Cycle.
Make the flow visible.
Find the bottleneck.
Understand why it is constrained.
Understand the behaviours and beliefs maintaining the system.
Identify the underlying failure mechanisms.
Decide what to fix first.

Then choose the appropriate tools.

That's the principle behind my approach to manufacturing improvement:

Diagnose before improving.

How to Diagnose Why a Factory Isn’t Performing