Analysis of internal customer response
Six Sigma – iSixSigma › Forums › Old Forums › General › Analysis of internal customer response
- This topic has 7 replies, 6 voices, and was last updated 13 years, 11 months ago by
DR J.
-
AuthorPosts
-
June 25, 2008 at 3:12 am #50392
We periodically carry out internal customer satisfaction survey.There are total 11 departments and the 150 respondents.The questionnaire having 10 questions each on prformance & behavioural aspects.The 5 point rating scale is used ranging form not at all to optimum level of meeting expectations.Till date we heve been carrying out simple analysis using average.Now I want to analyse the data using right statistical tool.I am confused whether I should use ANOVA or Correlation or both to give detailed feed back for actions.
0June 25, 2008 at 10:34 am #173181
SwaggertyParticipant@GeorgeInclude @George in your post and this person will
be notified via email.Well,
if it is subjective data, like strongly agree,agree, disagree, etc, you could use a likert scale
George0June 26, 2008 at 2:27 am #173230
Michael MeadParticipant@Michael-MeadInclude @Michael-Mead in your post and this person will
be notified via email.What you are talking about is a Likert scale.
Now, about your analysis. You do not have a big smaple, so a point estimate of your averages is not apporpriate. You need to put a confidence interval around your estimates to identify any real changes.
I have often seen statements like: “Last month the average was 3.09, and this month is 3.02. This can be attributed to a change in management.”
What a crock. This is hogwash. There is no reason to believe that this is natural variation. Of course the score will change from time to time, but is it significant? Even if it is not significant at the .05 level, you might be prompted to make some claim about the change. This can be reasonable if there is a trend in the statistics over time.0June 26, 2008 at 6:04 am #173232Break the responces into “Top 2 Box” (4 and 5 ratings), “Bottom 2 box” (1 and 2 ratings) and “neutral” (3 rating). Pie chart shall tell you the percentage contribution of each of the boxes. Obviously bottom 2 box needs to be enhanced.
DK0June 26, 2008 at 9:22 am #173236Dear Michael,
Your response is quite convencing to me.Can you help me to solve last part of my querry that should I use ANOVA or correlation or both.
Regards
MADAN0June 26, 2008 at 9:36 am #173238
Michael MeadParticipant@Michael-MeadInclude @Michael-Mead in your post and this person will
be notified via email.Hello Madan,
Sorry, I don’t know how to answer your question. I would do a t-test for the difference in means. Set a p-value of .10.
Another way would be to simply set a 90% confidence interval around your current estimate of the mean. If the new estimate falls within that range, conclude that there is not enough evidence to claim a change has taken place.
These tests can be done in Excel and many other statistics packages. I hope this helps. Maybe someone else can give you more information. Good luck.0June 30, 2008 at 1:40 pm #173358It’s a mistake to try to apply correlation in this case. ANOVA might be somewhat applicable but just barely. You have ordinal data you’re trying to analyze with “normal” techniques. The two shouldn’t be mixed in such a way. If you feel you must put numbers to these, you would be better off using non-parametric analysis. But the best way to analyze would be to simply chart the responses in a pie chart or bar chart. You still need to do the “heavy lifting” when it comes to understanding why you receive scores in the lower categories to understand what is going wrong (as well as in the higher categories to understand what you are doing right). Playing with ANOVA or correlation won’t help you in that respect.
0June 30, 2008 at 2:01 pm #173360Use ordinal or logistic regression so that you will know the causation between the customer overall score vs each behavior. This is critical in the analysis.
0 -
AuthorPosts
The forum ‘General’ is closed to new topics and replies.