A practical quality management handbook - clear, with examples and illustrations
How do you work faster and without errors? Lean teaches you to remove everything that adds no value for the customer. Six Sigma teaches you to remove errors and variation so that the result is the same high quality every time. This handbook walks you through the principles and tools step by step.
00 · Introduction
Every company wants to deliver faster, cheaper and without errors. Few, however, know exactly where their own processes lose time and money. That is the question answered by the two most widely used quality management methods - Lean and Six Sigma.
Both methods have a reputation for being complex fields full of statistics and Japanese terms. In reality, they rest on a few simple ideas. Once you understand them, you can improve processes without thick textbooks. This handbook therefore assumes no prior knowledge - it explains complex things in plain words, illustrations and examples, so that the owner of a small workshop, a production manager, a process engineer and a student writing a thesis can all take away the key insights of these methods.
You don't have to read it from cover to cover. It is written so that you can come back to it any time and quickly find what you need.
Lean was born at Toyota in post-war Japan, when the company could afford neither large warehouses nor unnecessary work. Taiichi Ohno and his colleagues learned to look at production through the customer's eyes: which of the things we do is the customer willing to pay for? Everything else - waiting, moving, inventory, rework - is waste.
Brings speed and simplicity
Lean chapterSix Sigma emerged decades later at Motorola in the USA. The company found that even a fast process is not enough if its results vary and a defective part occasionally turns up among the good ones. Six Sigma therefore measures the process, finds the causes of variation in the data and removes them.
Brings reliability and fact-based decisions
Six Sigma chapterThe two methods don't compete - they complement each other. Lean shows which steps are unnecessary; Six Sigma makes sure the necessary ones work equally well every time. That is why today they are most often used together - as Lean Six Sigma.
All chapters share the same structure, so you can find your way around quickly:
Find it in the table and jump straight to the tool that addresses it.
| If you are dealing with… | Use |
|---|---|
| Orders take too long, work keeps waiting for something | Value stream mapping, eliminating waste → |
| People keep searching for tools and material | 5S → |
| We produce to stock and the warehouse is overflowing | Pull, Kanban, Just in Time → |
| Quality varies, complaints and scrap are rising | DMAIC → |
| Too many problems, no idea where to start | Pareto chart → |
| We know what happens, but not why | Ishikawa and 5 Whys → |
| The line can't keep up even after adding people | Theory of Constraints (bottleneck) → |
| We are designing a new product or service | DMADV, QFD → |
| We want to prevent failures and risks | FMEA and Action Priority → |
| Strategy doesn't translate into daily work | Balanced Scorecard, strategy map → |
| We don't know how a change will play out over time | Dynamic simulation → |
A small company with 15 people takes orders for printed T-shirts. The process is simple: order → preparation → printing → drying → packing → shipping. Customers complain about long delivery times and the occasional smudged print. We will use this company to show how each tool works in practice.
01 · Lean
Lean looks for everything in a process that adds no value for the customer - waiting, moving, inventory, rework - and removes it. The result: shorter delivery times, less money tied up in stock, less stress.
You are cooking dinner. Only what ends up on the plate has value. Looking for spices, waiting for a free burner, or cooking twice as much and throwing half away - none of it makes the meal any better. Lean teaches you to spot such unnecessary steps and cook without them.
The biggest “aha moment” of Lean: most of the time, the order isn't being worked on - it is just waiting. In the queue for a machine, for approval, in the warehouse. That is why Lean focuses not on speeding up the work, but on shortening the waiting between it.
The first process map often shows exactly this picture. Printing a few minutes faster would hardly help PrintCo; cutting days of waiting would.
Work the customer won't pay for. A T-shirt waits two days for approval.
Fluctuating workload - a rush one day, nothing the next. 500 orders on Monday, 50 on Wednesday.
People and machines pushed beyond their capacity. The printer runs without maintenance breaks - and then breaks down.
Waste is any step the customer won't pay for. The acronym TIMWOODS helps you remember the eight types. Each comes with an example from our company:
Unnecessary moving of material and finished goods.
T-shirts are taken to an external warehouse and then brought back for packing.
More material and products than needed. They tie up money and space, and they age.
Thousands of T-shirts in colours that don't sell.
Unnecessary movement of people - searching, walking, reaching.
The printer walks to the far end of the hall for ink.
People or orders waiting for material, information, approval or a machine.
A finished design waits two days for the designer's approval.
Making more, or earlier, than the customer needs. It hides other problems.
Printing 300 pieces “just in case” when 200 were ordered.
Extra work the customer doesn't value - unnecessary precision, double checks.
Every T-shirt is packed in two bags.
Scrap, rework and complaints - the work is done twice.
Smudged print → T-shirt in the bin, print again.
People in the wrong roles or without room to suggest improvements.
An experienced printer knows how to speed up changeovers, but nobody asks her.
Automating a bad process only speeds up the waste. Remove the unnecessary steps first, and only then buy software or machines.
J. P. Womack and D. T. Jones described the approach as five consecutive steps. It is not a one-off project: after the fifth step you return to the first - with a better process than last time.
What does the customer really want and pay for?
The customer wants a quality print within 3 days - not a beautiful box.
Record every step from order to delivery and mark which ones add value and which don't. See value stream mapping (VSM) for how to do it.
The map shows 40 minutes of work and 5 days of waiting.
Remove waiting and queues between steps. Work in smaller batches.
The packing table moves right next to the printer and printed T-shirts are passed on continuously in batches of 20. Before, packing started only once the whole 500-piece order was finished.
Produce only against real demand - a customer order or a request from the next step, not a forecast for stock.
Printing happens only to order; the warehouse holds only blank T-shirts.
Improvement never ends. Involve everyone and repeat the cycle.
Every Friday for 15 minutes: what held us up this week?
Value stream mapping splits the journey of an order from order to delivery into consecutive blocks and records how long each one takes. A block is anything that happens to the order - work as well as waiting. For PrintCo, the map looks like this:
5 days in total, of which 40 minutes is work the customer pays for. The grey blocks are waste - and they are what is worth shortening. The timeline at the start of the chapter shows the same thing over time.
The map is best drawn on the shop floor, with pen and paper:
Push produces based on a forecast and hopes the goods will sell. PrintCo would print 200 T-shirts according to the latest trends and offer them in its online shop - whatever doesn't sell stays in the warehouse.
Pull starts production only on a real signal: a customer order, or a request from the next workstation that it is running out of material. There is less inventory and nothing is made unnecessarily. Where lead times are too short, pull is combined with a small, controlled buffer stock.
Material arrives exactly when it is needed - not earlier (inventory), not later (waiting).
A board with “to do - in progress - done” columns. At a glance, it shows where work piles up.
Small improvements every week from everyone - from operators to management. Small steps, big total.
Error-proofing: a part can't be fitted the wrong way, a machine won't start without the right setting.
02 · Six Sigma
Six Sigma uses data to find out why results vary and defects occur, and then removes the causes. It follows five steps called DMAIC. The goal is output you can rely on.
Target shooting: even if the average of all shots is dead centre, half of them can miss. The problem isn't the average, it's the spread. Six Sigma narrows the spread so that every result hits the target.
It isn't just for factories. Today, Six Sigma is used in healthcare, banking, logistics and software development - wherever an error costs money or customer trust.
The Greek letter σ (sigma) is the standard deviation - a number that tells you how much results vary. “Six sigma” means that the process is so stable that six such deviations fit between its mean and the tolerance limit. In practice: 3.4 defects per million opportunities.
To put it in perspective: a company at 3σ ships about 67 faulty parcels out of every thousand. At 6σ, it is one in roughly 300,000. Not every process needs 6σ - it makes sense where an error is very costly or threatens safety.
| Level | Defective per 1,000 shipments | Defects per million (DPMO) | Defect-free |
|---|---|---|---|
| 1σ | ≈ 691 | 691,462 | 30.85% |
| 2σ | ≈ 309 | 308,538 | 69.15% |
| 3σ | ≈ 67 | 66,807 | 93.32% |
| 4σ | ≈ 6 | 6,210 | 99.379% |
| 5σ | 1 in ~4,300 | 233 | 99.977% |
| 6σ | 1 in ~300,000 | 3.4 | 99.99966% |
A well-run company already collects data, documents procedures and improves. Six Sigma adds discipline: defects are removed systematically, decisions are based on numbers, and improvement follows a clear method with clear ownership.
It depends on whether you are improving something that already exists or designing something new. Choose what you are working on:
Five steps: Define → Measure → Analyze → Improve → Control. Each has a key question and continues the story of PrintCo and its smudged prints.
Describe the problem, the goal, the scope and what matters to the customer.
4% of T-shirts have a smudged print. Goal: below 1% within 3 months.
Measure the current state before changing anything. Check that you are measuring correctly.
For two weeks, we log every defect: type, machine, shift, time.
Find the root cause and confirm it with data - not with gut feeling.
Smudging is the most common defect and correlates with the dryer temperature.
Finding root causes: three tools anyone can use. No statistical software needed - just paper, a whiteboard and the team.
Rank the problems from the most frequent. You will usually find that a few causes create most of the trouble (the 80/20 rule). Focus your effort there.
Write the problem in the “head of the fish” and add possible causes along the bones in six areas: people, machine, method, material, measurement, environment. The tool stops you from fixating on the first idea.
It works best as a short team brainstorm at a whiteboard (15-30 minutes). The fishbone only collects suspects, though - the data decides which cause to really tackle.
Keep asking “why?” until you reach a cause you can remove. Five times is usually enough.
Design a solution, test it on a small scale and compare before/after.
Heater cleaning in the maintenance plan + a sensor that stops the belt when the temperature drops.
Set standards and monitoring so that the process doesn't slip back into old habits.
The temperature is logged automatically; defects stay below 1% in the long term.
Both methods address the same thing - designing a new product or service - and you pick one of them. They are supported by QFD and FMEA (see Practical tools).
What the customer needs from the new product and what the project goal is.
Turn wishes into measurable parameters (CTQs) - e.g. “the print survives 30 washes”.
Compare design options and their risks; verify them on a prototype.
Finalise the detailed design, test it and adjust it based on feedback.
Verify that the result meets the requirements and hand it over to operations.
The customer's needs.
Concept options and choosing the best one.
Fine-tuning parameters and tolerances at a reasonable cost.
Testing, validation, handover.
The roles are named after martial arts belts. The darker the belt, the more experience and responsibility:
Specific roles and certification requirements vary by company and certification body (e.g. ASQ, IASSC).
03 · Combined
Lean makes a process faster and simpler; Six Sigma makes it stable and precise. In practice, DMAIC serves as the framework, and Lean tools are used within its individual steps.
PrintCo would proceed like this: Lean removes the two-day wait for approval and the stock of finished T-shirts, and Six Sigma brings smudged prints below 1%.
| Lean | Six Sigma | Lean Six Sigma | |
|---|---|---|---|
| Asks | Where do we lose time and resources? | Why do results vary? | How do we deliver fast and defect-free? |
| Focus | Waste, flow, speed | Variation, defects, precision | Simplify first, then stabilise |
| Measures | Lead time, cycle time, inventory | DPMO, sigma level, Cp/Cpk | Both + financial benefit |
| Tools | VSM, 5S, Kanban, Kaizen, Poka-yoke | DMAIC, Pareto, Ishikawa, SPC, DOE, FMEA | DMAIC as the framework, Lean tools inside |
| Origin | Toyota, 1950s-1970s | Motorola, 1986 | Around 2000 |
04 · In practice
The tools used most often in Lean 6σ projects: from looking at the process through strategy and bottlenecks to risk assessment. Each one is explained so that you can use it tomorrow.
Every process has inputs (x) - people, material, machines, methods - and an output (y): a product or service. The output depends on the inputs: y = f(x). To get a better result, you need to find and influence the inputs that matter most. Feedback from the output then tells you what to adjust at the input.
Financial results show how the company did yesterday. The Balanced Scorecard (Kaplan and Norton, 1992) adds three more perspectives that show how it will do tomorrow. This turns strategy into specific goals and numbers:
For each of the four perspectives, define:
What we want to achieve.
How we will recognise progress.
How much, and by when.
What we will actually do about it.
A strategy map links the objectives of the four perspectives with cause → effect arrows. On a single page, it shows how investing in people and tools ultimately leads to profit - and everyone in the organisation can see how their work contributes to the whole.
In the example: better knowledge and tools make processes more efficient. That shortens cycle time and therefore customer waiting. Satisfied customers come back, revenue grows - and together with lower costs, so does profitability.
A chain is only as strong as its weakest link. A line produces only as much as its slowest step can handle - however fast the other steps are.
The Theory of Constraints (E. M. Goldratt, 1984) works in five steps:
The customer says “I want the print to last”. The manufacturer needs to know “how many washes, at what temperature”. Quality Function Deployment translates the voice of the customer (VoC) into measurable requirements and helps identify which of them matter most. Its best-known tool is the “House of Quality” matrix.
| The customer says | Technical requirement (CTQ) |
|---|---|
| “The print mustn't fade.” | Colour fastness after 30 washes at 40 °C |
| “I want it fast.” | Shipping within 48 hours of the order |
| “It should look exactly like my design.” | Print position deviation max. ± 3 mm |
For each process step, the team asks: What can go wrong? How bad would it be? How often does it happen? Will we notice it in time? The answers give the order in which to tackle the risks.
FMEA (Failure Mode and Effects Analysis) is a preventive tool. It shows where the critical points are - but you have to design the solutions yourself. It doesn't need precise data: the ratings come from experienced people on the shop floor, so it works even where numbers are missing.
For new and running processes: manufacturing, assembly, services, administration. Done during quality planning and used throughout operation.
During development of a new product, before production. By the end, all significant design weaknesses should be resolved.
This is a completed table for two steps of PrintCo's process. Each row is one risk. You fill it in from left to right - the header of each column contains the question you answer:
| 1Process stepWhich step is it? | 2Key inputWhat does the step depend on? | 3Failure modeWhat can go wrong? | 4EffectWhat does it do to the customer? | 5Severity (S)How bad is it? (1-10) | 6CauseWhy would it happen? | 7Occurrence (O)How often? (1-10) | 8Current controlsWhat prevents it today? (incl. SOP number) | 9Detection (D)Will we notice in time? (1-10) | 10Action Priority (AP)High / medium / low - from the AP table |
|---|---|---|---|---|---|---|---|---|---|
| Drying the print | Dryer temperature | Ink not fully cured | Smudged print, complaints | 7 | Clogged heating element | 5 | Visual check at the end of the line | 6 | Medium |
| Loading and printing | Placing the T-shirt in the printer | Misaligned print | T-shirt can't be sold | 6 | Manual loading without a stop | 4 | First-piece check | 4 | Low |
Download the FMEA templateCSV for Excel or Google Sheets · fill in AP using the table below · includes columns for actions and the new AP
You rate each risk with three numbers on a 1-10 scale and use a table to assign it an Action Priority (AP) - high, medium or low. Severity matters most, then occurrence, and finally detection. Try it in the calculator.
Action is required: improve prevention or detection, or document why the current controls are adequate.
General rating scales - each industry adapts them to its own needs.
| Effect | What it means | S |
|---|---|---|
| Hazardous without warning | Affects safe operation, with no warning | 10 |
| Hazardous with warning | Affects safe operation, with a warning | 9 |
| Very high | System inoperable, destructive failure without compromising safety | 8 |
| High | System inoperable with equipment damage | 7 |
| Moderate | System inoperable with minor damage | 6 |
| Low | System inoperable without damage | 5 |
| Very low | System operable with significant loss of performance | 4 |
| Minor | System operable with some loss of performance | 3 |
| Very minor | System operable with minimal interference | 2 |
| None | No effect | 1 |
The likelihood that a cause leads to the failure. Best estimated from historical data.
| Likelihood | Frequency | O |
|---|---|---|
| Very high - failure almost inevitable | ≥ 1 in 2 (≥ 50%) | 10 |
| 1 in 3 (33%) | 9 | |
| High - repeated failures | 1 in 8 (12.5%) | 8 |
| 1 in 20 (5%) | 7 | |
| Moderate - occasional failures | 1 in 80 (1.25%) | 6 |
| 1 in 400 (0.25%) | 5 | |
| 1 in 2,000 (0.05%) | 4 | |
| Low - relatively few failures | 1 in 15,000 (0.0067%) | 3 |
| 1 in 150,000 (0.00067%) | 2 | |
| Remote - failure unlikely | ≤ 1 in 1,500,000 (≤ 0.000067%) | 1 |
The harder the failure is to detect, the higher the number. At the start of a project, detection is usually rated fairly high.
| Detection | Chance that controls detect the failure | D |
|---|---|---|
| Absolute uncertainty | Controls cannot detect the failure | 10 |
| Very remote | Very remote chance of detection | 9 |
| Remote | Remote chance of detection | 8 |
| Very low | Very low chance of detection | 7 |
| Low | Low chance of detection | 6 |
| Moderate | Moderate chance of detection | 5 |
| Moderately high | Moderately high chance of detection | 4 |
| High | High chance of detection | 3 |
| Very high | Very high chance of detection | 2 |
| Almost certain | Controls will almost certainly detect the failure | 1 |
| Severity S | Occurrence O | D 7-10 | D 5-6 | D 2-4 | D 1 |
|---|---|---|---|---|---|
| 9-10 | 8-10 | High | High | High | High |
| 6-7 | High | High | High | High | |
| 4-5 | High | High | High | Medium | |
| 2-3 | High | Medium | Low | Low | |
| 1 | Low | Low | Low | Low | |
| 7-8 | 8-10 | High | High | High | High |
| 6-7 | High | High | High | Medium | |
| 4-5 | High | Medium | Medium | Medium | |
| 2-3 | Medium | Medium | Low | Low | |
| 1 | Low | Low | Low | Low | |
| 4-6 | 8-10 | High | High | Medium | Medium |
| 6-7 | Medium | Medium | Medium | Low | |
| 4-5 | Medium | Low | Low | Low | |
| 2-3 | Low | Low | Low | Low | |
| 1 | Low | Low | Low | Low | |
| 2-3 | 8-10 | Medium | Medium | Low | Low |
| 6-7 | Low | Low | Low | Low | |
| 4-5 | Low | Low | Low | Low | |
| 2-3 | Low | Low | Low | Low | |
| 1 | Low | Low | Low | Low | |
| 1 | 1-10 | Low | Low | Low | Low |
Priority used to be calculated as RPN = S × O × D (1-1000). Multiplication, however, weights all three factors equally: a risk of 10 × 2 × 5 has the same RPN (100) as 2 × 10 × 5, even though the first one is a safety issue. That is why the AIAG & VDA FMEA Handbook (2019), today's automotive standard, replaced RPN with the Action Priority table, which gives severity the greatest weight.
A bank assesses the risk that the amount is debited from the account, but no cash comes out. Very unpleasant for the customer - severity 9. The team found two causes:
| Cause | S | O | Current control | D | AP |
|---|---|---|---|---|---|
| ATM ran out of cash | 9 | 7 | Alert at the minimum cash level | 2 | High |
| Network error | 9 | 5 | None | 9 | High |
Both causes have high priority - action is required. Under the old RPN, running out of cash (126) would have looked like a much smaller problem than the network error (405); Action Priority correctly shows that with severity 9, both must be addressed. After the actions:
| Cause | Action | Owner | S | O | D | New AP |
|---|---|---|---|---|---|---|
| Ran out of cash | Higher minimum level at frequently used ATMs | Management | 9 | 4 | 1 | Medium |
| Network error | Higher network capacity + load balancing | IT + management | 9 | 2 | 3 | Low |
05 · Limits
The maps and calculations in this handbook work with averages or with a snapshot of the process at one moment. Real operations, however, change over time - and a planned process doesn't even exist yet.
The average queue in front of a machine can look fine - and yet the line stops twice per shift, because several slower orders arrive at once. In a running operation, data collection and measurement will reveal this, but only in hindsight: once the line has already stopped.
For a new process, layout or planned change, you have no data yet. This is where dynamic simulation helps: a digital model of the operation in which processes are “played out” over time and different variants can easily be compared - a different layout, number of machines, buffer sizes or shift pattern. You can verify a new design, a capacity expansion or DMAIC improvements before you invest in them.
Quick reference
Sources
How to cite this articleNepor, V. (2026). Lean Six Sigma: A Practical Quality Management Handbook. Vrealmatic. Available at: https://vrealmatic.com/lean-six-sigma
With dynamic simulation, we show how Lean 6σ measures affect your production or warehouse over time - without disrupting operations.
