SIX SIGMA: the beginning of the end

What the press is saying

A reader sent me this from the South China Morning Post:

“Karate-style corporate model grows flabby

Is it a statistical measure, a new economy management philosophy or a weird cult? Consult a Six Sigma practitioner about the phenomenon and you are likely to be met with a steely gaze and the laconic reply: ‘How long have you got?’ At that point you should back off or risk death by drowning in a deluge of corporate hogwash. The Six Sigma litany, dreamt up by Motorola in the 1980s, is rife with frightening references to ‘paradigm shifts’ and ‘systematic methodologies’.

Adding to the terror is the fact that its practitioners are ranked like martial arts experts. iSixSigma, one of the rare online sources with anything intelligible to say on the concept, explains the grading system. The term Black Belt means project leaders good at stats, the dreaded ‘interpersonal communication’ and well, Six Sigma.

Green Belts have had less training than Black Belts and take responsibility for leading fewer projects, while Master Black Belts spend almost all their time consulting, mentoring and training.

Just like his martial counterpart, a Six Sigma Black Belt is ‘self-assured and knowledgeable, the result of intensive training and real-world experience’. Better yet, he is ‘disciplined, purposeful and decisive, able to lead highly focused efforts aimed at improving a company’s bottom line,’ iSixSigma claims.

Above all, sources close to Technopedia aver, black belts drink coffee – lots of coffee – so that, like Napoleon Bonaparte, Catherine the Great and other renowned sociopaths, they never sleep.

Another human weakness they never indulge in is self-doubt. The typical personal potted history of a Six Sigma sensei reads like this: ‘Marc Zaphod is Senior Master Black Belt for Metahype – a recognised Six Sigma company. He is revered as a pioneer in the development of Six Sigma and has many lifetimes’ experience in applying process-improvement methodologies. He has personally trained more than eight million executives. His devotion to excellence has resulted in hundreds of millions of dollars in savings for corporations worldwide’.

This exalted position is achieved through TQM (Total Quality Management), which in turn breaks down into the equally totalitarian acronym DMAIC, which stands for Define, Measure, Analyse, Improve and Control.

According to the business skills development group Excel Partnership, any outfit prepared to submit to this regime should be guilty of only ‘3.4 defects per one million opportunities’.

That means holy Six Sigma status. Many companies apparently perform only at a paltry Three Sigma level, which means they are guilty of ‘roughly 67,000 defects per million opportunities’. Quite how Excel Partnership came up with either statistic is perplexing, but nobody would doubt their authenticity because funny, odd numbers exude authority.

Excel Partnership stresses that any firm which achieves the perfect six is rewarded by ‘dramatic improvements in business performance’.

As evangelists portray it, Six Sigma is an intensive training programme that can transform a weak and flabby business into a world-beater.

But Anand Sharma, chief executive of TBM Consulting Group and author of a Six Sigma critique called The Perfect Engine, views it differently.

‘When Six Sigma started appearing in annual reports of major corporations it seemed like a monument under construction that would change the face of business. However, more and more companies are tearing it off its pedestal,’ he told Technopedia.

Blame speed ‘or lack thereof’. Six Sigma programmes are falling out of fashion because they suffer from a lack of immediacy, Mr Sharma claimed, pointing the finger at the Black Belts ‘who run the show while other team members are relegated to being spectators’. Projects apparently drag on for months.

‘Six Sigma treats defects like a crime scene, testing from every conceivable angle, waiting months to receive ‘the answer’ from Black Belts, eventually delivering quality savings, but never quite managing real breakthrough improvements in lead time, productivity and inventory reduction,’ he said.

Perhaps it’s time for Six Sigma itself to evolve into something leaner and meaner. Seven Samurai, perhaps?”

Quite. Of all the bad press Six Sigma is getting, I think this is the best so far – the journalist has attitude. So now I turn to what is actually wrong with it.

The plausible sold to the gullible

How is Six Sigma sold? It will reduce defects, give you tools and techniques for doing so, ensure the problems solved are the ones you want solved and will tie all that up in a structured, systematic intervention. Everything that would have high appeal to command and control managers.

The problems are the wrong problems

Managers’ current conceptions of their problems are command and control conceptions. In service organisations this means things like: ‘how do I get my people to do more? How do I make my service levels’ and so on. It is the wrong way to conceptualise the problem because peoples’ performance is governed by their system. DMAIC – the systematic method for ‘problem-solving ’ – starts at plan, so managements’ current (wrong) pre-occupations define the problems. Change should start at ‘check’ – get knowledge about the ‘what and why’ of current performance as a system. There is nothing in Six Sigma that will help you do that. If you did do that you’d learn how management thinking and the way it translates to the design and management of work is the real problem.

You don’t need all that training

Just ask anyone who has been through all the training how much of it they use, let alone remember. The wholesale training of all and sundry to the various levels just represents great revenues for the training providers. More than that the training appeals to managers’ implicit models of change: training and projects. Not the ways to change a system.

Reporting distorts the system

The reporting requirements do what they always do in command and control systems – distort the system. So you can’t get anything done if you don’t call it a Six Sigma project and you report ‘results’ because it is required. I have sat next to ‘master black belts’ at their award ceremonies and been told the ‘results’ are not true by their colleagues, and it is of no surprise to me. I have regular correspondence from those who witness distortions in their system caused by Six Sigma reporting. We should expect it; it is in the nature of managing remotely with arbitrary measures.

A grain of truth

Like all management fads, Six Sigma has a grain of ‘truth’. It all started in Motorola. All they were doing was, in essence, Taguchi, reduction of variation, which is vital in improving product manufacturing. Having got good results in product terms the people involved decided to put the method into what they called an intervention framework (‘how do we get others to do this?’). Hence the command and control features of six sigma interventions (above). But the statistical approach morphed from Taguchi – more on that later as it is really only of interest to anoraks (and tricky stuff it is too).

But of greater importance right now: Service is not like manufacturing. Service organisations differ in that the customer is involved in production; there is inherently more variety in demand. To be successful (cut costs and improve service) service organisations need to be designed to absorb variety. Command and control designs just don’t do that, hence all the waste associated with poor service and high costs. Worse, when you specify a ‘six sigma problem’ in a command and control way, you damage the system’s ability to absorb variety; your Six Sigma project might look great but may actually be undermining performance (but you would not know).

The statistics are misleading

Six Sigma statistics start from the Six Sigma proposition: defects per million opportunities. Any ‘defect’ is defined by being beyond the specification. The Six Sigma statisticians use a ‘normal’ curve to determine this and their statistics are based on the relationship of that curve to the specification. They say the specification is customer-centric, but as I explain above you see specifications that are re-runs of managements’ current pre-occupations (for example, turning round customer application forms in five days).

Firstly, to my knowledge, no data are actually normally distributed; the normal distribution is a convention used as the basis for parametric statistics. Notwithstanding that, if you want to get data you can learn from you need data from the work, not data compared to an arbitrary specification (for example if time mattered to customers you would want to know the true end-to-end time from application to purchase – that’s the customer view – and those data should be time-series data as you can learn from variation).

The obsession with the bell curve is what led Jack Welch allegedly to insist that the bottom ten percent of employees be fired each year. Apparently his managers just hired temps to be seen to comply when the requirement for reporting getting rid of people came around.

And above all, if you use data in a bell curve, you cannot know if those observations falling at the tails are common cause or special cause variation. Six Sigma statistics will increase the probability that you act on common cause treating it as special cause – a sure way to sub-optimise the system; it increases variation. If Jack did this with his people he could be taken to court.

I told you it was tricky stuff! And if you want even more tricky stuff I can explain the madness of the ‘1.5 sigma shift’, but hey just take it from me: more smoke and mirrors.

TQM on steroids

Just like TQM it is training in tools and projects, but more training and a reporting structure. There is no requirement to change thinking and it is management thinking that is at the heart of our current problems.

Just as we saw with TQM it is bound to get some improvements with that level of investment and focus. But Six Sigma won’t get you the big prize. Just like TQM it will, in time, fail for it does not attack the real causes of sub-optimisation of performance.