Control charts
Six Sigma – iSixSigma › Forums › Old Forums › General › Control charts
 This topic has 11 replies, 8 voices, and was last updated 12 years, 5 months ago by Jonathon Andell.

AuthorPosts

November 23, 2009 at 5:12 pm #52959
Hi,I’m doing a CSAT project for a bpo. I’m little
confused with the result of my control chart and the
process capability of our process.
I took the weekly CSAT score(average score for the
week) for 25weeks(Number of surveys in each week
fluctaute). I used Xbar chart, the UCL = 1.114 and
lCL = 0.222(as per minitab).The mean is 0.446.The
process capability is 1.10. Does this mean my
process is stable? Although we have missed the
target of 52% many times in the 25 week sample. Am I doing something wrong? Can someone suggest if
Xbar chart is the right option in this case.
Please help me with the LCL and UCL calcualtion as
im confused with the formula used in minitab.0November 23, 2009 at 7:45 pm #187007Process capability has no relationship to process stability. Stable processes are those that contain commoncause only variation. This is evidenced by a control chart that depicts all data points randomly dispersed around the mean and within the control limits. If you are using continuous data, you should be using a Variability Chart with your Xbar Chart (ie R or S chart). You would want to confirm the variaility within your subgroups (ie R or S Chart) is in control before you can reliably assess the variability between your subgroups (X Chart). In terms of using a nonstandard subgroup size, you want to check into how that might effect your chart. Good luck.
0November 23, 2009 at 7:55 pm #187008abc,
Don’t confuse control and capability concepts as they are distinct concepts!
Control charts assess stability and whether, in this case, your 25 weekly CSAT readings represent a consistent process, with a mean of about 0.446. Check the chart. Does it show trending, either up or down, or any other evidence of nonstable behavior? Are there any outliers, obvious points above or below the control limits? If I assume correctly, and these CSAT values have a min of zero and a max of 1, then there is enough variation in these weekly readings to encompass your entire range. It sounds to me like you should be using an Individual XMR type chart based upon the inputs provided, so use that with the average weekly results, and see if the chart changes.
Capability refers to the ability of a process to meet limits which are typically imposed from outside, usually by a customer. In this case, that probably does not exist, and the capability indices are potentially meaningless. If artificial goals are applied, such as having a reading of 0.52, or higher, than the result should be phrased something like, “the data show X% of weekly readings were above the goal”. The analysis should then focus on what drives these readings to be higher, or lower, and assessing the reasons for low CSAT scores from customers.
0November 26, 2009 at 2:48 pm #187091
Jonathon AndellParticipant@JonathonAndell Include @JonathonAndell in your post and this person will
be notified via email.What is CSAT? Your sampling and subgrouping schemes may warrant further consideration – if you have significant sources of variation within your subgroups, the chosen chart may be hiding crucial information. Can you tell us a little more about your data?
0November 26, 2009 at 3:33 pm #187095
Ken FeldmanParticipant@Darth Include @Darth in your post and this person will
be notified via email.Customer Satisfaction. Have a good one Jonathon.
0November 26, 2009 at 5:41 pm #187099
Jonathon AndellParticipant@JonathonAndell Include @JonathonAndell in your post and this person will
be notified via email.Thanks for the acronym tutelage. I hope you are having a good year. It’s been a long time since I enjoyed working with you.Your response makes me wonder about the subgrouping scheme even more. Questions for the original poster to this discussion board:1. How many surveys per week?
2. What subgroup size do you use to compute control limits?
3. How did you establish a specification limit?Using continuousdata statistics with survey responses is a little tricky. I am not convinced that XBar is the best choice, but I am willing to be convinced.0November 26, 2009 at 6:48 pm #187100ABC – (I wish you could be a little more open than that, if only, with a surrogate name).I think you have oversimplified the development of your control charts and the conclusion. I also don’t understand how your center line will not be in the middle of UCL and LCL.
First, the number of observations you used to take average over each week should be the same. Not only the sample size should be the same, even the future test samples should be of the same size. If your sample is 310, you can choose the value of A2 from statistical control chart constants and set the limits at grand average + and – A2*Rbar, where Rbar is the range average. if n >10, you can use use std devn of the individual /sqrt(n) and then set the limits at + and – 3 std devns if you like. You should also develop range charts using statistical constants D3 and D4. For the chart average line and limits to be meaningful, you need at least 2530 weeks data – otherwise Normal distribution will not hold even for averages. In that case, you will need to use t distribution limits which are lot looser. These limits should also be subjected to process of reiteration until allout of control points are gone. If the process is unstable, i ahve found that many times we do not get the final limits at all as the process of iteration keeps removing new points either from the xbar chart or the range chart.10, you can use use std devn of the individual /sqrt(n) and then set the limits at + and – 3 std devns if you like. You should also develop range charts using statistical constants D3 and D4. For the chart average line and limits to be meaningful, you need at least 2530 weeks data – otherwise Normal distribution will not hold even for averages. In that case, you will need to use t distribution limits which are lot looser. These limits should also be subjected to process of reiteration until allout of control points are gone. If the process is unstable, i ahve found that many times we do not get the final limits at all as the process of iteration keeps removing new points either from the xbar chart or the range chart.10, you can use use std devn of the individual /sqrt(n) and then set the limits at + and – 3 std devns if you like. You should also develop range charts using statistical constants D3 and D4. For the chart average line and limits to be meaningful, you need at least 2530 weeks data – otherwise Normal distribution will not hold even for averages. In that case, you will need to use t distribution limits which are lot looser. These limits should also be subjected to process of reiteration until allout of control points are gone. If the process is unstable, i ahve found that many times we do not get the final limits at all as the process of iteration keeps removing new points either from the xbar chart or the range chart.Once the limits are set, you should use both (Xbar and R) charts to establish process stability over 8 weeks. You should not only watch for outliers but their are at least 6 other rules (e.g., 8 consecutive points on the same side of central line) that need to be observed. Finally, you also need to rule out trend, oscillation, and population mix ups – for 8 weeks (rule of Student’s t distribution) – to conclude process stability. Process capability is a different thing altogether which just establishes that the process is capable of delivering the specs/requirement.
Hope this helps.0November 27, 2009 at 6:43 am #187106Hi everyone,Thanks so much for your help.
We have five kinds of rating in our CSAT. Excellent,
very good, good, below average and poor. These
rating carries 2,1,0,1 and 2 points respetively.
To calculate the CSAT we multiply the survey with
its corresponding value and divide by total survey.
EX in a week we may get the following surveys: 5
Excellent, 7 Very good, 6 good, 2 below average and
2 poor.the CSAT is calculated as:
(5*2)+(7*1)+(6*0)+(2*1)+(2*2)/22. The average
score for the week would be 50%.We have surveys varying between 20 40 per week.My aim is to reduce the variation in CSAT score, my
findings sugest that its the SCR(single contact
resolution) that drives CSAT so i’ve suggested
measures to increase that which is related to the
process. I also want to know is there any hidden
factor that impact these scores, any trend that im
not able to notice. Im pretty new to this six sigma
concept so I’m wondering if Im using the right
tool(control chart) here.As far as the control limits are concerned I
calculated it using minitab and im still not
convinced as to how these valuse are calculated.
Would be great if any one of you can help me out
with that too.Thanks again for all your inputs.Abc.0November 27, 2009 at 10:38 am #187109greetings abc.i suggest that you have a one on one with your six sigma coach/es to better understand this concept. there’s a plethora of available materials with great content about control charts. but in my observation,
reading is not as much informative, interactive and fun when compared to a live discussion. the responses of the professionals here are helpful. but a forum has its limit – it’s like talking to
someone on a one way radio. to better help you with your data preparation, minitab has a pdf download that tells you what particular
control chart is applicable to your data. you can download it here.
http://www.minitab.com/enUS/training/tutorials/methodchooser.aspx0November 27, 2009 at 11:21 am #187111I don’t think the Xbar Chart is the right tool to monitor CSAT scores . YOu need to study the subgrouping carefully. You may like to use the XmR chart . Process Capability and stability are entirely different . Also pl check your data for normality – of late there’s been a war on this subject , thoughBBUSA
0November 27, 2009 at 1:24 pm #187114
Ken FeldmanParticipant@Darth Include @Darth in your post and this person will
be notified via email.ABC,
We have done this before in many applications. The easiest way is to define what you consider to be the best…that is, what would constitute a satisfactory score. We usually used “top two box” and then plotted a simple p chart using the number of surveys which received the top two ratings and divide that by total surveys. You are dealing with discrete data so taking averages, standard deviations and any other calculation based on continuous data is not right. In your case, I would have defined Excellent/Very Good as the top two box and thus had 12/22 or .545 as your proportion. Since the denominator will be varying week to week you can now simply plot a p chart. But, be real careful about reacting week to week. There are many variables contributing to CSAT so week to week variation is really meaningless.0November 27, 2009 at 4:13 pm #187117
Jonathon AndellParticipant@JonathonAndell Include @JonathonAndell in your post and this person will
be notified via email.abc, I am concerned whether this data fits with an XBar chart. There are a number of mismatches: – Your “range” can be only an integer from 0 to 4. There really should be more “shades of gray” available. Your “n” for subgroups makes me wonder whether you are using Rbar or S for setting control limits. Using this kind of scale to track variation (instead of mean responses) can be dicey at best.You may want to consider an IMR chart for the weekly averages. And as others have stated, don’t respond to a single week’s fluctuation unless the chart clearly indicates special cause variation. You may be unhappy with the wide limits an IMR chart will give you. However, the narrow limits of an RBar chart with “n” between 20 and 40 could be causing loads of false alarms.
0 
AuthorPosts
The forum ‘General’ is closed to new topics and replies.