Principle 6: Variability (Item 8)

All systems and processes exhibit variability, which impacts on predictability and performance.

Why this is important

The last several questions have made a strong case to reduce your variation. We will not repeat it. When you reduce your variation, you reduce your rework and all the costs associated with rework. For most companies, rework costs are significant.

Measuring consistency

How does your company know that it provides products and services at consistent quality to your customers? How do you measure what you provide? If you say "we ask our customers if they are satisfied", wrong answer! You must measure your processes that make a difference to your customers, before they get to you customers.

For example, most customers have a need for timeliness. You must constantly measure your response time. Even if your customer has never set a formal time. You must measure response time and be constantly reducing it – by identifying common causes and one-by-one eliminating them.

Customers also usually have need for accuracy. Measure your accuracy (or lack of it) with invoices and all transactions.

If you were a manufacturer of cement, your customers would have set specifications about the lime and silica content for different usages. You would have to monitor your processes to ensure you stayed within those specification limits. (For cement, these specifications are usually described in a `standard'.) That is, you must measure to make certain that you can produce within the customer's specifications every time. In terms of our discussion from Principle 4 (`To Improve the Outcome, Improve the System'), you measure make certain that your process is capable of delivering to these specification limits.

Error trapping

As you work to reduce your rework, you to work more and more at the errors that are not detectable by eye.

Most company's systems work at the easy-to-do level. At these levels, you can use methods like proofreading to find typing errors. They are OK when errors occur as often as a typo every page or so – one in a thousand. However, you cannot use the same methods when your errors are as infrequent as a typo per library.

The methods you use to trap errors by eye break down when you have many transactions. Most companies still use the same old paper based error trapping systems of 100 years ago. That is, proofreading.

Six Sigma

Six sigma is a methodology pioneered by Motorola. It is very useful for reducing the range of variation. Six sigma forces you to look at the different orders of magnitude that are necessary at each stage when reducing variation. Here are two examples.

If you have one typing error in every six words, that is one sigma. It would be easy to find. Two sigma is an error every four lines — also easy to find. Three sigma is a typo about every two pages – becoming more difficult. Four sigma is a typo every 70 pages. Five sigma is a typo every 20 books – a shelf full! Six sigma is a typo every 5,500 books – a library full! Most of our systems work at the easy-to-do level. At these levels, you can use methods like proofreading to find typing errors. They work to find errors that occur as often as three sigma – a typo every page or so. However, we cannot use the same method when our errors are so infrequent as a typo per library and we want to find and remove them. Who would care anyway?

Airline passengers do care when the errors are lost luggage rather than typos. Three sigma is one piece of luggage lost every two jumbos. Four sigma is one piece of luggage lost every 90 jumbos – a morning at LA airport. Five sigma is one piece of luggage lost every 10,000 jumbos. Six sigma is one piece of luggage lost every three million jumbos. Judging by the frequent use of lost luggage facilities, luggage handling systems work at about three or four sigma. Misplaced luggage (rather than lost forever) probably operates at one or two sigma.

It is clear from these examples that the solutions we use at one sigma level will not work at the next. Suppose we decide to reduce typos or lost luggage to the six sigma level. How would we go about it? Nothing we know about trapping or eliminating errors at our current level would help us at the next level down. How do we shift our thinking to the next level? Why would we want to?

Consider the banks. How many transactions a day does a big bank handle? Despite the huge transaction volume, most banks still operate at levels of two to three sigma – error trapping is done by eye. It is based on the old paper systems of 100 years ago. What is the risk to them of these errors? How much do they spend on error trapping and rework at this sigma level? What cost saving would they make if they could eliminate all that error trapping and move to the next level of sigma? Four sigma is an error every 31,000 transactions. Given the volumes of transactions involved, four sigma does not even sound extraordinary. It would be only an error every hour in most very large banks. However, it would need a huge change of thinking to even get to that level.

[Of course, `sigma' is based on the `sigma' used by statisticians to represent `variance' when they describe the normal distribution. The exact values are:]