Data Measurement

Quick reference

• 00:04 Hi, I'm Ray Sheen.
• 00:06 Measurement system analysis is all about trusting the data.
• 00:10 So let's start with a quick review of the need for data.
• 00:15 Data is often something that is measured.
• 00:17 The data may be used to make specific judgments
• 00:20 about the thing that is being measured.
• 00:22 That's what we do with the inspection data on a production line.
• 00:26 We decide if the item was built correctly because of some aspect that is measured.
• 00:31 Or we may be measuring something around us
• 00:34 to be able to make judgments about how it could have affected a problem.
• 00:38 For instance, we may measure the temperature or
• 00:40 the lighting to see if they contributed to an accident.
• 00:43 We'll also make measurements and create data in order to draw an inference about
• 00:48 something that the item that's being measured represents.
• 00:51 In manufacturing, we do this by taking samples from a batch of chemicals in
• 00:55 a pharmaceutical line to determine whether the entire batch is acceptable.
• 01:00 Or we may collect data from a consumer about their satisfaction
• 01:04 as part of an opinion poll, and use that to extrapolate marketing data for
• 01:07 the entire population.
• 01:10 Sometimes the data we measure is not about the item of interest,
• 01:13 but rather it is about the process.
• 01:16 This could include measuring and
• 01:17 collecting cycle time data for a manufacturing line or a call center.
• 01:22 It could be used to measure the uptime on an IT network or
• 01:25 to measure the ability of a sensor to detect color change.
• 01:29 Regardless of the type of measurement, data is created.
• 01:32 And that data is used to make business decisions.
• 01:36 It's not a great leap of logic to come to the conclusion that it's better to have
• 01:40 accurate data when making a decision than data that is unreliable.
• 01:44 Which brings us to the next point about data measurement system, an analysis of
• 01:49 these systems to make sure that the data is accurate and can be relied upon.
• 01:54 Let's focus the discussion a bit more.
• 01:57 When a company decides to improve the business, products or
• 01:59 processes, they want to ensure that the effort is effective.
• 02:03 That means they need to measure the results of what they have done.
• 02:06 The Lean Six Sigma methodology for problem solving and
• 02:09 process improvement relies heavily on the use of data.
• 02:12 That means it also relies heavily on a data measurement system.
• 02:17 When a data measurement system is detecting changes in the item that
• 02:20 is being measured,
• 02:21 there are always two causes potentially contributing to that change.
• 02:25 There is the actual change or variation of the item being measured.
• 02:29 There is also the variation or uncertainty in the measurement system.
• 02:33 Obviously, if the measurement system variation is too large
• 02:36 compared to the changes of the item being measured,
• 02:39 a wrong conclusion could be drawn about the item.
• 02:42 The uncertainty of the measurement system would overwhelm the changes it is being
• 02:46 asked to detect.
• 02:47 Which is why it is necessary to conduct a measurement system analysis.
• 02:51 This analysis determines whether the measurement system is appropriate for
• 02:55 the types of changes and the attributes that are being measured.
• 02:58 If the magnitude of variation of the measurement system is too large,
• 03:01 the resulting data measurement cannot be relied upon, and
• 03:05 the measurement system should be improved or replaced.
• 03:08 If the measurement system variation is small and
• 03:11 predictable, the measurement coming from the system can be trusted.
• 03:15 Let's explore the idea of measurement error a little bit more.
• 03:19 I'll start with one of the basic tenets of Lean Six Sigma.
• 03:22 All processes have variation in them, and the measuring of an item is a process.
• 03:28 And like with any type of variation,
• 03:30 if it is large it can impact the overall system performance.
• 03:34 That means that it could significantly change the measured value of an item
• 03:37 as compared to its true value.
• 03:39 And, again, it's no surprise that since this is a process with variation,
• 03:44 the process must be designed and managed well to minimize and
• 03:47 control the variation that comes from using the measurement system.
• 03:51 This diagram illustrates the point.
• 03:53 The triangle represents the measurement system.
• 03:56 The horizontal side is the standard deviation or
• 03:58 typical variation in the item being measured.
• 04:01 The vertical side of the triangle is the standard deviation of the measurement
• 04:05 system error.
• 04:06 The measurement is the value that is actually measured and recorded.
• 04:10 Using the Pythagorean Theorem, we see that the square of the measurement value
• 04:15 is equal to the sum of the squares of the true value of the item variation and
• 04:19 the measurement system variation and/or error.
• 04:22 If the measurement system error is small,
• 04:25 the measured value will be almost identical to the true value.
• 04:29 If the measurement system error grows,
• 04:31 it soon becomes a major contributor to the length of the hypotenuse.
• 04:35 And if it's too large as compared to the variation of the item that the system is
• 04:38 trying to detect, it will totally overwhelm the measurement, and
• 04:42 the data cannot be trusted.
• 04:44 I had this condition with a manufacturing company not long ago.
• 04:48 Their product required a precise alignment of some optics to work correctly.
• 04:52 They were having a lot of problems aligning those optics.
• 04:55 At times it would take many tries until they had acceptable performance.
• 04:59 Part of the alignment process included measuring the product at
• 05:03 a partially assembled state and then making final adjustments in the assembly.
• 05:07 We did a measurement system analysis and found that the measurement system
• 05:10 variation for this partially assembled alignment check
• 05:14 was ten times larger than the variation they were attempting to detect.
• 05:19 This alignment system was worthless for this application.
• 05:22 The Lean Six Sigma project that was charged with improving the product design
• 05:26 quickly changed to improve the measurement system.
• 05:29 The product was good, the measurement system was the problem.
• 05:33 Business relies on data.
• 05:36 Measuring business products and processes creates that data.
• 05:40 We need to have a measurement system that ensures that we have
• 05:44 accurate data available to us.

Lesson notes are only available for subscribers.