# converting likert mean VOC surveys to 6 sigma?

Six Sigma – iSixSigma Forums Old Forums General converting likert mean VOC surveys to 6 sigma?

Viewing 11 posts - 1 through 11 (of 11 total)
• Author
Posts
• #29663

mcintosh
Participant

How do most of you, who use six sigma, use customer satisfaction surveys? Specifically, how do you define a customer defect on your survey?For instance if overall satisfaction overall, or on a process overall is asked on a 7 point scale (1=not at all satisfies to 7=very satisfied), do you count an answer of 5 as 2 defects and convert to DPMO then a sigma score? Or is it more appropriate to set a limit of say 6, and then any customer rating a 5 or below, is a defect in the process?

0
#76467

T
Member

0
#76468

Participant

I’ve never done this type of calculation myself, but if I were to suggestion something….
I don’t think it’s appropriate to say that you receive 2 defects for someone giving you a 5 out of 7 on the scale. It’s like saying that it’s a defect if a customer doesn’t think you provided outstanding service on that particular topic. You may want to do that, but I think it exceeds that point of diminishing returns. Instead, I would think that anything below satisfactory is a defect (not the customer is a defect, as you wrote on your previous email).
So let’s say that I’m the customer taking your survey. I read “how well did company XYZ do in delivering the product on time, as you requested?” Assuming a customer can respond 1, 2, 3, 4, 5, 6, or 7, if I gave you a 1, 2, or 3 it would be a single defect. This one question would be the opportunity. You could then have other questions focusing on the other CTQs of the customer (those too would be opportunities).
You would want to capture why a customer gave you a 1-3 so you could analyze the defects to determine root cause and eliminate. You would also want to trend your responses over time because a 4 rating probably isn’t good enough either.
The other point I wanted to make was that this is a trailing indicator. It does not really tell you how your process is doing with enough time to change it. You could receive this customer feedback 2 weeks after you delivered a product, meanwhile, 2 weeks worth of other customers may also not be having their CTQs met. You want to move trailing indicators to leading indicators, maybe through the use of SPC.
Hopefully this helps. Good luck.

0
#76473

Mike Carnell
Participant

Tom,
I think the two defects for a 5 is bit of self flagalation. Probably won’t make your data mean any more than what it does if you count it as one.
There was a study, white paper, etc floating around that had shown that a one and a five (ina 5 point scale) were the only numbers that reflected a customer feeling strongly enough to take some form of action – tell afriend it is great or pick up the phone and complain. I think it was around GE. Maybe some of those people have seen it?
Good luck.

0
#76522

Harjot Singh
Participant

Converting Likert to Six Sigma is a tough task as it is … and while the following might not always apply, you could:
On a 7 point scale, assume anything below 6 is a defect (or below 5 or 4) … depending on your LSL (which is customer defined; customer being internal or external). Define opportunities and defects robustly. Then count the individual defects and opportunities and use the DPMO methodology. It should help!
The basic caveat is that since you are attempting to convert a discrete metric into a continuous metric, it will always be rough on the edges.
Best of luck with the solution and let me know if it worked.

0
#76525

Mark Almeter
Participant

One thing we never want to lose sight of is “What is the data telling us”?  Our VOC survey process employs a likert scale with “Top 2 Box satisfaction” being the measurement, and “Satisfied” as the benchmark.  Our process is also designed to capture as many comments as possible, with further probing if necessary (i.e…If dissatisfied, what is the single most important reason why you answered the way you did?).  We Pareto all comments, and focus on the critical dissatisfiers.  The top 2 satisfaction measurement is trended over time.  Thus, our process is designed to maximize use of critical information through a blend of both quantitative and qualitative methods.
We have to remember that VOC surveys are an information gathering tool, and measuring VOC using the “defect” method and trying to fit the VOC scores into Six Sigma metric terminology can defeat the purpose of what you’re ultimately trying to achieve.

Hope this helps….
Mark

0
#76534

Chantale
Participant

Hello,You need to identify what your target is first. If a score of 6 and above is your target, then everything below 6 would be considered as a defect. It’s as simple as that.Have a good day!

0
#76548

Ellyn Brognara
Participant

We use a 5 point scale – very satisfied, satisfied, neutral, disatisfied and very disatisfied. We calculate defects as bottom 2 box – any response of disatisfied and very disatisfied.

0
#76549

Marty Martin
Participant

Customer surveys are very subjective. Looking at your data for on-time delivery, returns and customer complaints would provide facts, “I know” rather than “I think”. Applicable metrics for on-time shipments and rejections will allow you to calculate DPMO and Sigma.
Customer surveys may help you understand that the customer is not satisfied but you have no idea why they are unsatisfied. If you do understand the 5 W’s you still need to research the facts to make sure it’s not customer perception.
Customer perception and satisfaction are key drivers for any business. Talking to the customer, tracking customer complaints and understanding delivery and quality performance issues will also give you great direction for process improvement.

0
#76598

Stephen Curtis
Member

You could use a 4 point Likert Scale (no centre-point, forced decision) and score all that are “strongly disatissfied” & “somewhat disatissified” as being defects, ie something that you don’t want to see.  This can then be treated as attribute data.

0
#76601

Mike Carnell
Participant

Tom,
I haven’t been on this thread in a while but you need to really think about the advice Mark Almeter gave you. There are probably as many “right” statistical answers as there are people to send them.
It will always come back to 2 basic questions that you need to set up before you go out and collect data:
1. What do I want to know?
2. How do I want to see it?
The question should lead you to what your question actually is for the customer. That should help you develop a hypothesis which should help you understand the data you need and how you want to analize it.