Lean Six Sigma is a combined improvement method that pairs lean's attack on waste and slow flow with Six Sigma's attack on variation and defects, run through one structured problem-solving cycle called DMAIC. Lean makes the work fast; Six Sigma makes it consistent. Together they fix both.
That is the whole idea in a paragraph. The rest of this guide covers where the two systems came from, what the "sigma" in Six Sigma actually measures, how a DMAIC project runs step by step, when to lead with lean and when to lead with Six Sigma, what the colored belts really mean, and how the combined method differs from a single tool like statistical process control. This is the methodology overview; the individual tools each get their own guide.
What is Lean Six Sigma?
Lean Six Sigma is a disciplined method for improving a process by removing two different problems at once: waste, which lean targets, and variation, which Six Sigma targets. It is not two programs bolted together for a poster. It is one project structure, DMAIC, that reaches for lean tools or statistical tools depending on which problem the data in front of you says is bigger.
The reason to combine them is that the two problems hide each other. A fast process full of variation produces defects faster. A statistically perfect process buried under excess inventory and handoffs still takes six weeks to deliver a two-day job. Lean without Six Sigma cleans up flow but leaves a wobbly process that drifts back out of spec. Six Sigma without lean tightens one step while the value stream around it stays clogged. You need both lenses, and ASQ defines the combined method precisely because most real improvement needs aspects of each.
Put simply: lean asks "why does this take so long and touch so many hands?" Six Sigma asks "why is the output different every time?" Lean Six Sigma asks both in the same project and refuses to declare victory until flow is quick and the result is repeatable.
Where did lean and Six Sigma come from?
The two halves have separate origins and met later. Lean grew out of the Toyota Production System, built principally by Taiichi Ohno at Toyota from the late 1940s onward and named "lean" by researcher John Krafcik in 1988. Its whole aim is eliminating the eight wastes and letting value flow at the pull of the customer, the story told in full in our guide to lean manufacturing.
Six Sigma is younger and American. Engineer Bill Smith developed it at Motorola in 1986, and Motorola made it a company-wide program to drive out defects using statistics. Motorola later reported saving billions over the following decades by running it relentlessly. Where lean came from the plant floor and the stopwatch, Six Sigma came from the quality lab and the control chart, it leaned heavily on the SPC methods Walter Shewhart had invented at Bell Labs in the 1920s.
By the late 1990s companies running both noticed they were solving different halves of the same problem, and the combined label "Lean Six Sigma" stuck. The marriage is practical, not theoretical: a lean team that hit a variation wall borrowed statistics; a Six Sigma team drowning in a slow value stream borrowed flow tools. Neither system had to change to cooperate.
What does the "sigma" in Six Sigma actually mean?
Sigma is the Greek letter statisticians use for standard deviation, a measure of spread. A process running at "six sigma" is so consistent that six standard deviations fit between its average and the nearest specification limit, which works out to just 3.4 defects per million opportunities once you allow for the long-term drift real processes show. That is the origin of the name and the target.
The useful part is the ladder. Each sigma level corresponds to a defect rate, and moving up a level is dramatically harder and dramatically more valuable than the last. Most plants that have never measured sit around three to four sigma and are shocked when they convert their scrap rate into the table below.
| Sigma level | Defects per million opportunities (DPMO) | Yield | Rough real-world read |
|---|---|---|---|
| 2σ | ~308,500 | ~69.1% | Uncompetitive; heavy rework |
| 3σ | ~66,800 | ~93.3% | Common for un-measured processes |
| 4σ | ~6,210 | ~99.38% | Typical "we're pretty good" plant |
| 5σ | ~233 | ~99.977% | Best-in-class for most industries |
| 6σ | 3.4 | ~99.99966% | Rare; required in aerospace/medical critical features |
You do not need to hit six sigma everywhere. The point of the ladder is direction and honesty: it converts a vague "quality is fine" into a number you can move, and it shows why a jump from 4σ to 5σ earns far more than the sigma value suggests, you cut defects roughly 25-fold. Six Sigma projects are usually justified on exactly that math, straight into your cost of quality.
| Fact | Detail | Source |
|---|---|---|
| Six Sigma coined | By engineer Bill Smith at Motorola, 1986 | ASQ |
| Six-sigma performance | 3.4 defects per million opportunities | ASQ |
| Core project cycle | DMAIC: define, measure, analyze, improve, control | ASQ |
| Practitioner roles | White, yellow, green, black, and master black belts | ASQ, Six Sigma belts |
| Lean's five principles | Value, value stream, flow, pull, perfection | Lean Enterprise Institute |
What is DMAIC and how does a project run?
DMAIC is the five-phase cycle every Lean Six Sigma project follows: define, measure, analyze, improve, control. It is the shared backbone that lets lean tools and statistical tools live in one project, you use whichever fits the phase you are in. The discipline is finishing each phase before the next; teams that jump to "improve" before they have measured almost always fix the wrong thing.
- Define. State the problem, the customer requirement, and the goal in one page. What is defective, how often, and what would "fixed" look like as a number? A vague charter guarantees a vague result. Map the process at a high level and confirm the pain is real, not anecdotal.
- Measure. Establish the baseline with trustworthy data. Prove the gauge first, if two people measuring the same thing disagree, every later number is fiction. Then collect enough real data to state today's defect rate or lead time honestly, so you can prove the fix later.
- Analyze. Find the root cause, not the loudest symptom. This is where the two toolkits meet: a lean team maps the value stream and hunts the eight wastes; a Six Sigma team runs a Pareto, tests hypotheses, or designs an experiment to isolate what actually drives the variation. Confirm the cause with data before you touch anything.
- Improve. Change the process to remove the confirmed cause, then verify the change actually moved the baseline. This is where kaizen events mistake-proofing, flow redesigns, and setting-optimization live. Pilot before you scale.
- Control. Lock the gain in so it cannot slip back. Update standard work put the key output on a control chart with a reaction plan, and hand ownership to the process owner. A project without a control phase is a loan: the metric drifts back the month after the team disbands.
Lean or Six Sigma: which do you lead with?
Lead with lean when the problem is speed, handoffs, or clutter; lead with Six Sigma when the problem is a defect that appears at random and nobody can predict. Most plants should lead with lean first, because waste is usually visible, cheap to remove, and removing it exposes the variation problems that were hiding underneath.
| If the symptom is… | Lead with | Because |
|---|---|---|
| Long lead time, work sitting in queues | Lean | It is a flow problem; map the value stream |
| Excess inventory, motion, waiting | Lean | Classic waste; see the three enemies muda, mura, muri |
| Defect rate that swings unpredictably | Six Sigma | It is a variation problem; needs statistics |
| Process drifts out of spec over a shift | Six Sigma | Needs SPC and capability study |
| Both: slow AND inconsistent | Lean first, then Six Sigma | Remove waste, then attack the variation it exposed |
What do the belts mean, and do you need them?
The colored belts describe how much Lean Six Sigma training and project experience a person has, borrowed from martial arts. ASQ recognizes white, yellow, green, black, and master black belts. A green belt runs improvement projects part-time alongside a day job; a black belt leads bigger projects full-time and coaches green belts; a master black belt trains and mentors across the site.
You do not need a belt to improve a process, and treating certification as the goal is a common way for programs to stall, you end up with credentialed people and unchanged metrics. Belts are useful as a shared language and a way to build internal capability, not as a trophy. A plant that has removed real waste with metrics that moved is further along than one with a wall of certificates and no change on the floor. Start with a real problem, not a training calendar.
How is Lean Six Sigma different from SPC?
Statistical process control is one tool that lives mostly inside the control phase of DMAIC; Lean Six Sigma is the whole method around it. SPC answers a single question, is this process stable and predictable?, with a control chart. That is essential, but it does not define the problem, remove waste, find the root cause, or redesign the flow. Those are the other four DMAIC phases.
Think of it as scope. If someone says "we run SPC," they have a chart watching one characteristic. If they say "we run Lean Six Sigma," they have a repeatable way to take any process problem from vague complaint to locked-in fix, and SPC is the instrument they reach for in the last phase to make sure the fix holds. Both matter; they are not the same size. Our SPC guide covers the chart in depth, and process capability covers the "is it good enough" question that follows stability.
Where does Lean Six Sigma pay off first?
It pays off first on a process that is both slow and inconsistent, where the cost of the current mess is measurable, and where you can get honest data on the baseline. Pick a value stream customers complain about, put a number on today's lead time and defect rate, and run one full DMAIC loop on it before starting a second. Resist the urge to launch fifteen projects; a single finished project that moved a real number builds more belief than a program office full of half-done charters.
The most common failure is not statistical, it is data. Teams cannot establish a trustworthy baseline in the measure phase, or cannot prove in the control phase that the gain held, because the plant's numbers live on paper logs totaled at month-end. That is a plumbing problem, and it is worth fixing before the first project, not during it.
What does Lean Six Sigma look like with real-time data?
The method predates cheap sensors, and every phase runs faster when the data is live instead of hand-tallied. Measure gets honest when the baseline comes from the machines instead of clipboards. Analyze gets quick when you can slice defects by shift, line, and material lot in seconds instead of retyping logs. Control actually holds when the key output sits on a live chart that pages someone the moment it drifts, rather than surfacing in next month's report.
This is the layer Harmony provides: digitize the data capture operators already do, connect the machines and the systems you already run, and compute true numbers from the source, with no rip-and-replace. When CLS moved production logging off paper, supervisors went from finding problems in the next morning's report to seeing them during the shift they happened, which is exactly the feedback speed a DMAIC control phase needs. You can see how the data layer fits the tools you already use.
Start with one slow, inconsistent process. Run one honest DMAIC loop. Make sure you can measure the baseline and prove the gain daily. That order, not the belts and not the software, is what separates plants where Lean Six Sigma works from plants where it is a binder on a shelf.