Pros and Cons of Six Sigma: Where It Pays and Where It Fails
Is Six Sigma worth implementing? Yes, when a process repeats often, its defects cost real money and the data already exists; no when the work is a one-off, the cause is an obvious broken step or the real aim is invention. Almost every disadvantage people list sits on the second side of that line: training time, slow projects, paperwork and a narrow view of quality. This guide prices those costs and shows how organisations reduce them.
Six Sigma grew at Motorola in the 1980s and then spread to other manufacturers and service firms. Its pitch is simple: pick a process, measure its defects, find the causes with data and fix them for good. The pitch is also where the trouble starts, because each word in it (pick, measure, data, for good) is a place where an implementation can go wrong. The old argument of "pros versus cons" is therefore less useful than a sharper question: under which conditions do the pros appear, and what does it cost to create those conditions?
Where does Six Sigma earn its keep?
Six Sigma pays when four conditions hold together. The process runs often enough to produce plenty of observations. Each defect has a price someone can name, such as rework, scrap, refunds or a missed delivery. One person owns the process and can change it. And the cause is not obvious, so analysis adds something that a quick fix would not.
When those conditions are in place, the method brings real strengths that older improvement drives often lacked:
- Money is attached to every project. A project is approved only after its effect on the bottom line has been defined, which stops improvement work drifting into pet topics.
- Decisions rest on data. Teams must measure before they argue, which settles many disputes that opinion never would.
- Roles are defined. Champions, Master Black Belts, Black Belts and Green Belts give a project leaders, coaches and a route for escalation.
- Human and process sides are handled together. Customer focus and culture change sit alongside process control and statistical analysis, rather than being left to chance.
- Tools have an order. Each technique belongs to a phase of the project, so a team does not reach for a design of experiments when a simple chart would show the answer.
A separate article on the measured benefits covers how each gain is counted. Here the question is when those gains stay out of reach. The table matches common situations against the method.
| Situation | Fit with Six Sigma | Why |
|---|---|---|
| Thousands of invoices a month with an error log | Strong | Plenty of counts, a visible cost per error and a single process owner |
| Delivery delays recorded with timestamps | Strong | Variation can be charted and the causes tested |
| A yearly audit that runs once | Weak | One observation a year cannot separate pattern from chance |
| A form that lacks a required field | Overkill | The cause is plain, so correcting the form beats a months-long project |
| A process with no records at all | Not yet | The first project is building the measurement, not improving the result |
| A new service concept | Weak | There is no running process to measure, so design methods fit better |
What do the downsides actually cost in time and money?
The costs are real and they arrive before the savings do. Four of them come up in almost every honest review of a rollout.
Belt time and training fees
A Green Belt keeps a day job while running a project, so the cost is mostly capacity taken from other work. A worked example shows the scale. The figures are illustrative assumptions except the exam fee, which is the published price of the LSSA Green Belt exam.
| Cost line | Working | Amount |
|---|---|---|
| Project time | 10 belts x 1 day a week x 17 weeks | 170 person-days |
| Value of that time | 170 days at an assumed €300 a day | €51,000 |
| Exam fees | 10 x €532 | €5,320 |
| Subtotal before course fees | €51,000 + €5,320 | €56,320 |
| Break-even per project | €56,320 divided by 10 projects | About €5,632 of net, validated savings |
The course fees from a training provider come on top. The point of the arithmetic is the last row: unless each project returns that much after costs, the cohort has taught people a method at the company's expense.
Slow projects
Many belt projects run for several months, because the Measure and Analyse phases need data to accumulate before anyone can draw a conclusion. A manager who wants a result next week will find that frustrating. The usual answer is not to rush the analysis but to pick problems whose data already exists, so the calendar is spent on analysis and not on waiting for numbers.
Bureaucracy and the wrong scoreboard
Six Sigma slides into box-ticking when leaders count the number of trained Black Belts and Green Belts or the number of projects closed, instead of the money kept. People respond to what is counted: they open easy projects, split large ones into small ones and close them early. Templates and review meetings multiply, and the time spent reporting on improvement starts to exceed the time spent improving.
Overselling and uneven credentials
The belt system created a training industry, and critics argue that some consulting firms claim expertise in Six Sigma while understanding only a little of the tools or of the client's industry. Credentials vary as well. Wikipedia's article on Six Sigma notes that there is no single certification body, that criteria for Green and Black Belts differ, and that some companies ask only for a course and a project. A belt from one source may therefore mean something quite different from a belt from another.
The same article summarises a literature review that found 34 common failure factors across 56 papers on Lean and Six Sigma published between 1995 and 2013. The leading groups were weak management commitment, a lack of training and resources, poor project selection and weak links to strategy, and resistance to cultural change. None of those is a flaw in the statistics. They are flaws in how the programme is run.
Why does a data-poor process stall a Six Sigma project?
Six Sigma assumes that you can count what goes wrong. When no one has recorded defects, the first months of a project go on building a way to record them, and that stage can take the largest share of the calendar. The team has promised savings and has so far produced a spreadsheet.
Three situations cause most of the delay:
- Nothing is recorded. A back-office team handles requests by email and has no log. Someone must sample requests by hand for several weeks before analysis can begin.
- The definition is slippery. A "late" delivery means late against the promise to the customer in one department and late against the warehouse plan in another. Until the team agrees one definition, every chart is open to challenge.
- The measurement itself is unreliable. Two inspectors classify the same item differently, so part of the variation the team sees is the measuring, not the process. A measurement system check, which compares repeated readings by different people, exposes this before the team chases a cause that does not exist.
The practical answer is to treat the data gap as the project. A first project that only sets up a clean count of defects is a good, cheap use of a Green Belt, and it makes the second project fast. A wasteful alternative is to guess a baseline, run a full DMAIC cycle on it and report a saving that nobody can defend later.
Is the sigma number a fair score for service work?
The headline claim of Six Sigma is a process that makes 3.4 defects per million opportunities. That figure rests on a convention, and critics have attacked it from two sides.
The 1.5 sigma shift
According to Wikipedia, practitioners add an empirical 1.5 sigma shift to allow for the fact that processes drift over the long term. A process that fits six standard deviations between its mean and the nearest limit in a short study is then treated as a 4.5 sigma process in the long run, and 4.5 sigma corresponds to the familiar 3.4 per million. Without the shift, a six sigma process would produce about two defects per billion. The shift is therefore a large assumption applied to every process, and a small error in it changes the calculated defect level. Whenever you quote a sigma level, say which convention you used, or quote the defect rate directly.
All defects are not equal
A raw defect count treats every miss as the same, which service work rarely supports. Take a hospital admissions desk that handles 2,000 admissions a month and logs two kinds of error. The team agrees weights in advance: a mistyped address scores 1 and a missing allergy flag scores 20. These figures are an illustration, not data from a real unit.
| Error type | Weight | Before | After | Weighted before | Weighted after |
|---|---|---|---|---|---|
| Mistyped address | 1 | 60 | 20 | 60 | 20 |
| Missing allergy flag | 20 | 4 | 4 | 80 | 80 |
| Total | 64 | 24 | 140 | 100 |
The plain count fell by 62.5 percent (64 to 24), which makes a good slide. The weighted score fell by 29 percent (140 to 100) because the dangerous error did not move at all. A team that reports only the count has declared victory while the serious risk remains. The lesson is to define defects in terms of customer harm before any measuring begins, and to track the severe ones separately.
Targets that may be too loose or too tight
The quality expert Philip Crosby argued that the standard was too lenient for complex products: a chip with millions of circuits, built to 3.4 defects per million opportunities, would still contain flaws. That objection matters less for an invoicing process and more for anything where one failure is catastrophic. Where a single defect is unacceptable, a sigma target is the wrong tool and engineered safeguards come first.
Does tight process control squeeze out innovation?
Partly, and the honest answer depends on what kind of work is being controlled. Six Sigma is built to reduce variation. That is exactly right for a billing run and exactly wrong for a research lab, where useful surprises are the product.
Wikipedia reports a BusinessWeek account of Six Sigma at 3M, where its introduction was described as having stifled creativity and the method was removed from the research function. Two business school professors are cited as arguing that it leads to incremental innovation at the expense of blue-sky research. Whether or not you accept every detail, the pattern is plausible: when scientists must justify steps and measurements, they try fewer things.
Organisations that avoid the trap usually do three things:
- Keep the method on operations. Order handling, service delivery, claims and production get DMAIC. Early-stage research and concept work do not.
- Use the design branch for new products. Design for Six Sigma, known as DMADV, applies the same measuring habit to a design before it exists, so customer requirements are defined first and the design is verified with pilot runs.
- Protect the exploring time. A team that is asked to hit a defect target every quarter has no slack for experiments. Leaders who want both set aside time that has no project attached.
There is a related complaint that Six Sigma is not original. Joseph Juran called it a basic version of quality improvement and said it contained little that was new. That does not make it ineffective, but it does mean that a company which already practises disciplined quality management will gain less than one starting from nothing.
How do organisations soften the drawbacks?
Companies that keep Six Sigma for years rarely run it in its pure form. They adapt it in six ways.
1. Combine it with Lean
Lean looks at flow and waste: waiting, rework, handoffs and extra steps. Six Sigma looks at variation. Lean Six Sigma treats them as complementary, which has a practical payoff. Lean often removes the obvious waste in a few days, so the slow statistical work is saved for problems that really need it.
2. Scope each project tightly
A project that tries to fix an entire end-to-end process drags on and loses its sponsor. One process step, one owner and one measurable target is far more likely to finish. Scope that is too narrow can be widened in a second project, while scope that is too wide is almost never narrowed once the team has started.
3. Pilot before a company-wide launch
Train a small cohort, run two or three projects in one function and compare the programme cost with the validated savings after a few months. If the pilot returns more than it consumed, expand. If it does not, the company learns that cheaply and can change course before hundreds of staff are trained.
4. Score projects before approving them
Choosing projects by pure judgement is a known weak spot, and it is one reason programmes stall. A simple screen is better than none. Score each candidate from 1 to 5 on the size of the prize, the readiness of the data, the likely speed of the fix and the strength of the process owner, then add up. The scores below are illustrative.
| Candidate project | Prize | Data ready | Speed | Owner | Total | Decision |
|---|---|---|---|---|---|---|
| Invoice errors | 4 | 5 | 4 | 4 | 17 | Start now |
| Warehouse mis-picks | 4 | 4 | 4 | 5 | 17 | Start now |
| Call handling time | 3 | 5 | 4 | 3 | 15 | Start now |
| Customer onboarding delays | 5 | 2 | 2 | 3 | 12 | Measure first |
| New product concept | 5 | 1 | 1 | 2 | 9 | Use a design method |
A cut-off of 15 is arbitrary, but any agreed cut-off beats a debate. The onboarding project is valuable and the data is poor, so its first step is the measurement project described earlier.
5. Reward money kept, not belts produced
Report net savings after programme costs, and have the process owner confirm that the gain held. Drop leaderboards of certified staff. This one change removes much of the bureaucratic drift.
6. Standardise what a belt means
If belts come from different providers, define internally what a Green Belt must be able to do, such as complete a project, show the data and present the result. An external exam gives a common baseline for the knowledge part, while the project proves the practical part.
When is it better to fix the process another way?
Run a candidate through these questions before committing a belt for four months:
- Is the cause already obvious? If yes, fix it this week and measure the result.
- Does the process produce enough events to chart? If it runs a handful of times a year, use a checklist and a review instead.
- Can you name the cost of a defect in money or customer harm? If not, the project has no business case yet.
- Does one person own the process? If not, appoint an owner first, because a project without one cannot make the change stick.
- Is the goal to improve something that exists, or to invent something new? Invention calls for design and research methods.
A project that survives all five is a good candidate. One that fails two or more should wait, shrink or use a lighter method. Saying no to a poor candidate is itself a mark of a mature programme, and it protects the credibility of the projects that go ahead.
How does this link to the LSSA Lean Six Sigma Green Belt exam?
The Green Belt is the role that carries out the scoped projects described above, so a certificate in it is a sensible way to show you understand both the tools and their limits. The LSSA Lean Six Sigma Green Belt exam (exam code LSSA-GB) combines the Lean and the Six Sigma halves in one test.
| Exam detail | LSSA-GB |
|---|---|
| Full name | LSSA Lean Six Sigma - Green Belt |
| Exam code | LSSA-GB |
| Vendor | LSSA / Lean Six Sigma Academy |
| Number of questions | 60 |
| Duration | 180 minutes |
| Exam fee | €532 (LSSA web shop) |
| Passing score | 38 / 60 |
LSSA's own shop page for the Lean Six Sigma Green Belt theoretical exam describes it as 60 multiple-choice questions; check it for the current price and booking steps before you register. The syllabus is laid out on the LSSA-GB syllabus page, and it is worth reading with the trade-offs above in mind, because a candidate who can say when a tool should not be used usually answers scenario questions better.
To rehearse the 60-question format before booking, the LSSA-GB practice exam gives you a way to try it in advance.
Frequently Asked Questions
What are the main disadvantages of Six Sigma?
The main disadvantages are the cost of belt training and project time, slow projects when data is missing, a drift into paperwork when leaders count belts instead of savings, and a narrow focus on reducing variation that can suit routine processes better than creative work. Careful scoping and pilots reduce most of them.
Is Six Sigma only suitable for large manufacturers?
No. Service teams use it too, but it needs a process that repeats often and produces data. A small business with few repeated transactions will usually do better with simple Lean methods, a checklist or a quick root cause review before investing in belt training.
Does Six Sigma work for innovation?
Not in its standard improvement form. DMAIC is built to cut variation in existing processes, which can suppress exploratory research. For new products, organisations use the design branch known as DMADV, and they keep early-stage research outside the defect-reduction targets.
How many questions are on the LSSA Green Belt exam?
The LSSA-GB exam has 60 questions and allows 180 minutes. Candidates need 38 of the 60 correct to pass. LSSA's web shop lists the exam fee at €532, so confirm the current price on LSSA's shop page before you book.
What is the passing score for LSSA-GB?
The passing score for the LSSA Lean Six Sigma Green Belt exam is 38 out of 60 questions. That is roughly 63 percent. Because the exam allows 180 minutes, most of the preparation effort goes into understanding when each tool applies, not into speed.
