Lean Six Sigma Green Belt – Practice Project Part 2

3. Practice Project_Part 3

So project leader, he wants to find out whether the potential inputs that have been shortlisted are critical or not. He wants to ascertain whether their inputs really affect the percentage of late payments that were translating a few of the payments into defectives or not. So he starts is collecting the data for all the potential inputs. We are going to look into that data set and we are going to mention the results and inferences of the hypothesis testing for each of those inputs. Let me take you to that minitab file and then we’ll solve it. Now, here is the minitab and here is the data for all the potential inputs that we have identified here.

Tax or turnaround time with tax, turnaround time without tax issue. So if I take you all here, you see that people are speaking about tax issues. Tax issues are critical, they take a lot of time to close. So we are trying to quantitatively evaluate and whether this is true or not. For this to be true, there has to be a difference between turnaround time with the tax issue and turnaround time without tax issue in order to prove that this is potential. Here we have the data since both you are defective or non zero one and the tax issue.

Yes or no? Both are discrete in nature. So your output variable is discrete and your input variables are also discrete. Hence we go with two proportion test. Now, if you have any confusion, I would request you to go back to the hypothesis testing of analyze phase and try to relook at two proportion tests and Kaisquatus, because those are the two tests which we are going to use predominantly.

Given this data and these inputs, even before we go ahead with the proportion test, we need to write the null and alternate hypothesis. Null hypothesis would be both the proportions are equal, that is proportion of defectives when it comes to turnaround time with tax issue and proportion of defectives when it comes to turnaround time without tax issues.

These two should be equal. So I’ll write it as proportion of A is equal to proportion of B. All right? And when it comes to alternate hypothesis, proportion of defective when it comes to tat with the tax issue is not equal to proportion of defectives when it comes to turnaround time without tax issues. So it will be proportion of A, sorry, is not equal to proportion of B, a being turnaround time with tax issue and B being turnaround time without tax issue.

In order to solve this, we need to do a two proportion test. So we need to go to stat basic statistics to proportion test. Here, since both these samples are in different columns, we need to do a drop down and select each sample is in its own column here we’re going to select turnaround time with tax issue and turnaround time without tax issue. And if I do the hypothesis test we get the p value which is zero and we know for a fact that p low.

If p value is less than alpha which is 0. 5, then p is less than 0. 5 implies p lo null go. Hence we reject null hypothesis which says that or if you reject null hypothesis, obviously you want to go with alternate hypothesis which says both the proportions are not equal. Hence, this input is critical for us because the number of defectives vary based on whether there is tax issue or not. So this happens to be a critical input for us. Let us look into the next input which is priority. If I go to ishikawa diagram, someone was saying about the priority. Here we go. I guess the priority of invoice affects how much time we take to close the invoice. So we need to determine on whether priority actually impacts the time to close the invoice or not. So we have the priority which is discrete.

Defective or not is also discrete. Since we have greater than two categories, low, medium and high, we need to go with chi square and within chi square we know for a fact that null hypothesis says all proportions are equal. Alternate hypothesis says not all proportions are equal. If you are not clear, I will request you to go to analyze phase and relook into the chi square test. We are now trying to implement our learnings so it will be a little faster. Just bear with me on that. Even before we apply chisquare, we need to arrange these three columns into two columns.

That is called as stacking the data. So I would go to data stack columns and I’m going to select these three priorities. Click and select I would say, hey, place these in the same sheet. We have C eight and C nine. So I would say let the subscripts be in maybe C eight and let the data be in C nine. Now if I click on OK, we have these two columns.

Now that we have stacked the data, we can perform a chisquare test. Let me go to stat tables, crosstabulation and chisquare here. All the rows row entries are 10, zero, so on and so forth. All those are in C nine. So I’m going to select C nine there and all these column names prior to Low, Medium and High are there in this column called C eight. And when you select that, I click on Chi square and I want to perform a chi square test. Click on OK and okay, there we go. We have the Pvalue here. Since this Pvalue is greater than 0. 5, that is Pvalue is greater than alpha, which means that the Pvalue is greater than 0. 5, which implies p Hi null fly.

So if you fail to reject null hypothesis which says that the proportion of defective across the priorities is the same, that means this input is not critical for us since the time taken or classifying a specific invoice as defective or not is the same across the priorities. So this is not critical. So let me go back here and say this is not critical. And we’ve already calculated and we found that whether there is a tax issue or not is going to determine on the criticality.

All right, now I’ll go back to the Mini tab. Let us look into customer type here. Customer type is corporate, individual or government. If I take you back to the inputs that we have identified using ishikawa diagram here, we need to check yes. So government customers need a million follow ups. That’s the major issue. So we need to evaluate and check whether these inputs are critical to us or not. So let me open this Minitab file and let us try to solve.

So these are those values, customer type, corporate, individual and government. These are all defectives. One means it’s defective, zero means it’s not a defective. That’s discrete. Customer type, corporate, individual or government is also discrete. So since we have more than two categories, we go with chi square test. Before that, we need to stack the data. That is how Minitab understands when it comes to chi square test. So I go to data stack the various columns. I’m going to select these three customer types and click on select there. There we go. And I wish to place the data in C 14 and C 13. And I click on OK, there we go.

We have the customer type and we have the row details that is defective or not. In C 40 I need to perform or we need to now perform chi square. So I go to stat tables, cross tabulation and chi square. We need to change these values because row details are there in C 14. I’m going to select C 14 there. Let me remove everything and then select C 14. And when it comes to columns, column details, which is customer type, corporate, individual and government is there in C 13 here. So I’ll select C 13. Initially, we have checked this option Kaiser for the previous test, hence that option remains perfect. Let me click on okay, now P value happens to be zero, so Pvalue is less than 0.

5, which implies that we reject null hypothesis p lo NALGO. So you reject meaning we say that p lo NALGO, that is we are rejecting null hypothesis, which implies that all proportions are equal. That means the number of defectives does not depend on the customer type or the consumer type. Hence, we will say that this also turned out to be non critical. All right, now let me go back to the Minitab file and let us now try to analyze the next input which is correct customer address or incorrect customer address. If I go back to the bunch of inputs which we have identified here, it says the major issue has been incorrect customer details on invoices. So we need to find out whether this is critical or not. Incorrect Customer Details let us try to solve this. Since we have two categories which are discrete, we have data on whether a specific invoice is defective or not, which is also discrete. We will go ahead with two proportion test. We go to stat basic statistics. Two proportion Test since each sample isn’t in its own column, I will remove whatever is existing here.

And we need to enter the new data here, which happens to be correct Customer Address. And then we also have incorrect customer address. Perfect. Now let me click on OK here. All right. P value happens to be once again less than 0. 5. So you reject null hypothesis and null hypothesis says both the proportion of defects are the same. That is, proportion of defects for correct customer address and incorrect customer address happens to be the same is what our test says. So once again, this input is non critical. Let us go back to the minitab file and let us try to understand what are the other inputs available. So we have geographies. Once again, I go back to the bunch of inputs that we have tried to identify here. Someone was saying that depending on the location, things change. There we go. So I did notice that it takes longer for certain customer regions.

 So we need to find out whether this is true or not. For that, let me go back here. Since we have three categories and since both X and Y are discrete, we need to stack the data. I go to data stack columns. We need to select Asia, Europe and us. And here we need to provide C 23 and C 22. Let me click on okay, there. There we go. Now we need to provide the chi square. We need to perform chi square test. I go to stat tables, cross tabulation and chi square. We need to select C 23 and 24. So I’m going to select C 23 here and we have C 24. Let me remove that and select C 22. There we go. Let me check whether chi square test is checked or not. Yes, it’s checked.

Let me click on by default it is not checked. Friends, you need to check that since we have checked it initially in this session, that particular option is selected. Now let me click on okay, there we go. So the p value is high. What does that mean? Pi nullify. The Pvalue is greater than 0. 5, which implies we fail to reject null hypothesis p hi null fly. So that means all the proportions or the proportion of defectives across the regions vary. They are not equal. Hence, this input is critical for our analysis. Let me type in that this is critical. I’ll go back to the presentation. We have the products right? So let me go back to the inputs which we have identified as part of a Chicago or Fishbone diagram. Yes, different product type. Invoices do have varying duration to close an invoice.

Is that true or not? We need to find out. Now, let me go back to the Mini tab. Since we have three categories, both are discrete, both X and Y. I mean, we need to stack the data first. Let me go to data stack columns this time. I’m going to select the three products and C. Let us place those in C 28 and 27. Here we go. Now, we need to perform chi squares by going to start tables, cross tabulation and chi square. I need to change this values to C 28 and C 27. Let me click on OK. Now wow here. Also, the Pvalue is greater than 0. 05, which implies you fail to reject null Hypothesis, which once again proves that this statement is true, that the number of defectives vary based on the product type.

I think we have one last input to deal with. That is our analysts aren’t experienced to deal with all situations. Is that true or not? Let us go back to the Minita. Finally, solve one last input here. Since we have two categories and both ynx are discrete in nature, we are going to perform two proportion tests. Go to Stat Basic Statistics on my back and click on two proportions. We need to remove the things which are here. And let me select the Analyst Experience greater than two years. Analyst Experience less than two Years well, now when I click on okay, it gives a Pvalue as 0. 46. Once again, Pvalue is greater than 0. 5, which implies that P Hi null fly. So we fail to reject null Hypothesis, which says that the proportion of defectives vary based on the Analyst Experience, thereby making the Analyst Experience input critical. So we have four critical.

4. Practice Project_Part 4

Let us now discuss about how to identify and shortlist the vital few courses from the trivial menu. So the project leader, he has also collected the type of courses found in each defective invoice and collected it in an Excel sheet. And we are expected to find out what are the vital few causes that should be reduced to reduce the number of defectives. Let us try to solve this using minitab. Here we go. We have the data here. We have all the issues and the number of issues that we have is approximately equal to 16,000. If you scroll down, there will be 16,000 such increase.

When we look into the count of each and every issue, we get the count here, the deficit count. I do not want to focus on all of these seven issues and try to address each and every issue. It doesn’t warrant the amount of effort that we spend. Hence, we need to find out what are those vital few costs out of these. For this we need to do something called as parado chart, which is also called as 80 20 rule.

All we need to do is go to stat quality tools and select parado chart in defects or attribute data. We need to select the issue type and when it comes to the frequencies, we have the frequencies in C four column which is defect count. All you need to do is that and click on OK. And this would generate a plot for Parato chart. We are still waiting for the result. It is coming up. There we go. So here we have the pareto chart.

If you look at these values closely, pending clarification issues contribute to 28. 2% of the overall issues. Delayed follow up would contribute to 21. 3% of the overall issues. Improper follow ups would account to 18. 2% of the overall issues. And then we have incorrect invoice which accounts to 12. 8% of the overall issues. But if you look into the first four issues, the cumulative percentage we get 80. 6%.

What does that mean? This means that out of the seven issues that we have, if we can focus on these four issues, we will be able to reduce the number of defectives to a significant extent. Let us go back to the PowerPoint and update these causes or these issues. There the critical not exercise or this critical issue types. The first critical issue type would be pending clarification.

The second critical issue type is delayed follow up. Third one is improper follow ups. And finally we have incorrect invoice. These four happen to be providing for us. Okay, let me arrange this. There we go. All right. Let us continue further.