Leaderboard

While continuous data is generally preferred over discrete data, please indicate circumstances where discrete is the preferred data type although continuous data is available for the same characteristic. Note for website visitors - Two questions are asked every week on this platform. One on Tuesday and the other on Friday. All questions so far can be seen here - https://www.benchmarksixsigma.com/forum/lean-six-sigma-business-excellence-questions/ Please visit the forum home page at https://www.benchmarksixsigma.com/forum/ to respond to the latest question open till the next Tuesday/ Friday evening 5 PM as per Indian Standard Time. The best answer is always shown at the top among responses and the author finds honorable mention in our Business Excellence dictionary at https://www.benchmarksixsigma.com/forum/business-excellence-dictionary-glossary/ along with the related term.

What would an excellence practitioner lose if he does not utilize the concept of rational subgrouping in the pursuit of process improvement? The principle underlying the concept of Rational Sub-grouping As per the Central Limit Theorem, the distribution of sample averages taken from a population will be normal distribution. The sample mean value of the sample averages will equal to the population average and the standard deviation of the sample averages will be σ / √n, where σ is the population standard deviation and n is the sample size. This principle is used for deriving the control limits for a control plan. By Rational sub-grouping, we mean samples taken in succession during a particular time. Usually the number of samples in a rational sub-group (i.e. the sample size) will be very small, say, 4 or 5. The next such sample has to be taken after a time interval. The reason for taking the samples in succession is to ensure that they will (predominantly) have only variations due to chance causes, since they are produced under very similar conditions. The reason to keep the sample size small is to minimize any assignable variations that could creep in due to too much time gap between samples. The below table gives a representation of how data may be organized in sub-groups. What if an excellence practitioner does not utilize the concept of rational subgrouping? Let’s consider the following possibilities, instead of picking up the rationalized sub-group as explained above. 1. If he picks ups one large set of samples with no sub-groups: Using such a sample, he will be able to prepare a frequency diagram with class intervals and study the characteristics such as mean and overall variance. The two types of variation, i.e. due to chance causes and assignable causes will be combined and he would not be able to distinguish them separately. He will not be able to construct a control chart to assess the different types of variabilities. 2. If he picks up sub-groups with large no. of samples in each sub-group: Each sub-group is likely to exhibit variations other than chance causes. This can magnify the range and widen the control limits, if a control chart is constructed using this data. This will reduce the sensitivity of the control chart to detect instabilities. 3. If he picks up the samples for the sub-group with larger time interval: Any variability due to special causes that could have happened between the intervals could be missed out. The causes that lead to any drift of mean value or expansion of variation (range) could get unnoticed. This could impact the correctness of the control limits derived. 4. If he does not give sufficient intervals between picking up each sub-group: The conditions of samples in one subgroup are likely to overlap with that of adjacent sub-group, depriving the practitioner from obtaining a realistic ‘between’ subgroup variation. This could result in reduced R values and lead to narrower control limits. 5. If he picks up just one (or two) sample each time: In the case of picking just one sample, the range will not get estimated and there will be no possibility of working out the control limits. In the case of picking just 2 samples, he is at risk of narrowing the range and hence the control limits. Thus, by not using the concept of rational sub-grouping, practitioner will fail to come up with the best assessment of the 3 types of variabilities viz Instability, Off-target, and Variation the existing process

About Baseline One of the requirements of the Measure Phase in Six Sigma DMAIC cycle is the Baseline measurement, sometimes expressed as Baseline Sigma. In fact it is hard to tell whether the baseline data is required as part of the Define phase or Measure phase. Ideally, if we need to give the problem statement, which is expected to cover What, When, Magnitude and Impact. The ‘When’ portion is expected to show the metrics related to the problem for a time period as a trend chart, so that we can see the magnitude of the problem and the variation over a period of time – and acts as a baseline. Baseline certainly helps to act as reference to compare and assess the extent of improvement. Baseline is important to get a good measure of the quantum of improvement and in turn to quantify the benefits in tangible terms. However, the following discussion brings out certain practical challenges related to Baseline. 1. Baseline metric did not exist, but is it worth post-creating it? Suppose we are trying to improve an electronic product, based on certain customer complaints, our project objective will be to ensure that the incidents of customer complaints should be reduced or eliminated. Upon subjecting the product to a special lab evaluation, we could simulate the failure. However, a reasonable baseline metric will be possible only if we subject a set of sample units for a certain period of time. This could prove quite costly and time consuming. On the other hand the solution to the problem is known and we may proceed with the actions. Since our goal is to ensure zero failure, under the given conditions and duration, comparison with a baseline is not important here. Many a time, when the company is anxious to implement the improvement to get the desired benefits, be It cost or Quality, it may not make much sense to build up a baseline data, unless, it is readily available. 2. New measurement methodology evolved as part of improvement Let’s take an example of Insurance Claims processing, where the payment / denial decisions are taken based on a set of rules and associated calculations. The improvement being sought is to reduce the rate of processing errors. However it was only as part of the improvement actions that an appropriate assessment tool was evolved to identify and quantify the errors by the processors. By this time, the improvement has already begun and it is not practically possible to trace backwards to use this tool and get a baseline measurement. 3. When improvement is for ‘Delight factors’ Often we introduce enhancement features on product, for example, new models / variants of smart phones. In such cases, the emphasis is more on the delight factors for customers, for features that they haven’t experienced earlier and any baseline comparison may not have much relevance. 4. Integrated set of modifications Let’s examine another scenario where a series of modifications were implemented on a software application and was released together as a new version. Here, the set of actions taken influenced multiple factors, including performance improvement, elimination of bugs and inclusion of new innovative features. In such situations, any comparison with a baseline performance to the current will be very difficult and would have overlapping impacts. If we still need to do a comparison before vs after, we may have to do so after factoring and adjusting for such interaction effects on the pre / post improvement outcomes. To conclude, in general, a baseline metric is an important information that we require to compare the post improvement results – However, it has to be borne in mind that certain situations challenge the feasibility and relevance of using a baseline measurement.