Process Stability

Quick reference

• 00:05 Hi, I'm Ray Sheen.
• 00:06 I wanna introduce the concept of process stability.
• 00:09 Understanding this concept will be a big help in the Analyze and
• 00:13 Improve phases of a Lean Six Sigma project.
• 00:16 Stability starts with an understanding of the categories of variation.
• 00:21 There are fundamentally two kinds of causes of variation.
• 00:24 The first is common cause variation.
• 00:28 It is the random variation that exists in any physical product or process.
• 00:32 When the only type of variation that is present is variation from common causes,
• 00:36 the process is stable and predictable.
• 00:39 This type of variation can be measured and modeled with statistics.
• 00:44 Thanks to this modeling, the variation can be predicted.
• 00:47 We know what to expect.
• 00:49 Designing your process to accommodate the level of
• 00:52 common cause variation will result in a very stable process.
• 00:56 Even though there may be minor common cause variation, you've accounted for
• 01:00 it in the process design and control.
• 01:03 The other type of variation is from special causes.
• 01:06 When special cause variation is present, we say the process is unstable.
• 01:11 These are causes that are unusual situations,
• 01:14 not part of the normal process performance.
• 01:17 In fact, they're often caused by the lack of process controls.
• 01:20 The management team or process operator decides to do something
• 01:24 different within the process, and thus tampers with the process performance.
• 01:29 This type of variation cannot be modeled with statistical
• 01:32 tools because it is not predictable.
• 01:35 We cannot predict when an unusual event will occur.
• 01:38 In order for a process to become stable,
• 01:41 the special cause variations must be identified and removed.
• 01:46 Let's look closer at common cause variation.
• 01:49 Common cause variation is always present in a physical process.
• 01:53 The process designers and process operators should expect it and
• 01:57 should allow for that level of variation within the process tolerances,
• 02:01 either the input tolerances or output tolerances.
• 02:04 Common cause variation is random in nature, and therefore,
• 02:07 will often be modeled with the normal distribution, the bell-shaped curve.
• 02:11 The precise values of the variation cannot be predicted for
• 02:15 any unique instance of the process.
• 02:17 But we can predict the average amount of variation, and
• 02:21 measure the standard deviation of that variation distribution.
• 02:25 A key point to remember is that it is always there,
• 02:28 which means we can't control common cause variation by taking special action.
• 02:33 We can't start trying to tweak the process as we chase a common cause variation that
• 02:39 is slightly higher or slightly lower than the last instance of running the process.
• 02:43 If we do that, we will inevitably make the system unstable.
• 02:47 We'll begin to over control.
• 02:49 Our tampering with that normal process operation becomes a source of special
• 02:53 cause variation because we try to manage the process differently on every instance.
• 02:59 So let's now look a little deeper into special cause variation.
• 03:03 When there is a special cause variation, a process is not stable.
• 03:07 By that we mean that we can't predict the amount of variability in this
• 03:11 process output.
• 03:13 The normal process controls do not react adequately to special cause variation
• 03:18 because it is outside what the process planners had envisioned would occur.
• 03:23 You may be able to react to the special cause variation but
• 03:26 only by taking a special action such as reworking the product or
• 03:30 scrapping the item that was impacted.
• 03:32 The point is that the item doesn't follow normal flow to successful result when
• 03:37 special cause occurs.
• 03:39 Special Cause is not random and actually, that's a good thing.
• 03:43 It means we could track down the root cause, and
• 03:46 it's usually a singular root cause.
• 03:49 Knowing the root cause, a special control or action can be taken to eliminate that
• 03:53 root cause and thus eliminate the effect that it has had on the process.
• 03:58 But one caution with that, sometimes the special cause creates a special good
• 04:03 process output that cannot be repeated by the normal process control and management.
• 04:08 Now this often happens when everyone is watching closely
• 04:11 as we try a new process step or introducing a new product.
• 04:15 There's a lot of extra attention and help at each step in the process, but
• 04:19 when that attention goes away, the process cannot sustain the performance.
• 04:24 This is referred to as the Hawthorne Effect,
• 04:27 based upon some research that was done 100 years ago.
• 04:29 That research was trying to determine the effect of improving working conditions
• 04:33 on productivity at a plant in Hawthorne, Illinois.
• 04:36 It turned out that the most important factor for
• 04:38 improving productivity was not the working conditions, rather was all the special
• 04:43 attention the workers were receiving because of the study.
• 04:47 So let's summarize this discussion with a head to head comparison.
• 04:50 Common cause variation is the normal variation in the process and
• 04:55 it is random in nature.
• 04:56 It is inherent in the system design, the equipment, the materials, the procedures,
• 05:00 the operator skills and training and the measurement in controls.
• 05:04 Because it is a function of the design, it can be modeled and
• 05:08 the levels of variation are predictable.
• 05:11 It cannot be avoided or controlled without fundamentally changing the process.
• 05:16 So when it comes to improvements, if you only have common cause variation,
• 05:21 you'll need to be recommending a significant change to a new process
• 05:24 that inherently has less variation.
• 05:28 But we contrast that with special cause variation.
• 05:31 It is unpredictable, which also means it is not random.
• 05:35 There is something that has occurred that is outside the normal
• 05:38 control of the process, something that was not accounted for in the process design.
• 05:43 The occurrence cannot be mathematically predicted.
• 05:46 We can't calculate the damage after the fact but we can predict when or
• 05:50 what will happen.
• 05:52 Overall, a good thing is that we can often put controls in place to screen out
• 05:57 the special causes and
• 05:58 therefore we don't need to invest in creating a brand new process.
• 06:05 A key goal in the Analyze phase will be to determine if the variation in
• 06:09 the process is due to special cause or common cause.
• 06:13 Eliminating the special cause will stabilize the process and
• 06:17 allow you to put it under predictable statistical control.

Lesson notes are only available for subscribers.