# Subjective Measurement Plan

Six Sigma – iSixSigma Forums Old Forums General Subjective Measurement Plan

Viewing 15 posts - 1 through 15 (of 15 total)
• Author
Posts
• #36732

Cannizzo
Participant

Has anyone had to create a measurement plan for something subjective?  I am working with my team on creating a measurement plan, and one aspect of it is subjective – Clean.  It’s pretty relative.  How do I set up a plan to determine if something is clean per our requirements?
Thanks!

0
#106688

Heebeegeebee BB
Participant

Well,
What is the part?
What are the Customer Specs?   (Have you run a QFD??)
-Have you talked to your Customer?  Better yet, invite the End-User to be a part of your team and drill them for UDE’s/DE’s.
Application? (Aerospace, commercial, toys, etc…)
IF Aerspace, or medical, you could start with Cleanroom requirements (i.e. class 10K, class 100K)
And a potential measure could be particulate remanants at final (PPM)
It really depends on what the Customer really wants.
We need more data.

0
#106690

Cannizzo
Participant

We’re currently completing our CCR’s, which are based on customer surveys.  This is for construction equipment – the Clean factor applies to an entire machine (not a part).  So it’s comparable to how clean a car is.  Customer wants the machine clean, but not sparkling for the most part.  So how do I define the CCR’s to be measurable?

0
#106691

miranax
Participant

There are several techniques for assessing so called “subjective criteria”.  Two that I can think of right now are Attribute Gage RR’s and non parametric Techniques such as the Kendall Coefficient of Concordance and the Spearman Rank Correlation Coefficient (the Quality Council of Indiana CQE Primer has information on the non parametric techniques).  In any case Heebeegeebee is right: first you need to come up with a operational definition of what “clean” is.  Once you get a parameter in which you can make comparisons, I would suggest you use the Attribute Gage RR of the Minitab Software.

0
#106693

Heebeegeebee BB
Participant

I’d recommend a full-blown QFD focusing on getting the Customer to define and establish specific, measureable KQC’s/KPIV’s.
Pin your Customer down to Specific definitions that are measureable.
Also, I’d try a SIPOC and Criticality Tree.

0
#106694

Heebeegeebee BB
Participant

But how would an Attribute GR&R address cleanliness with out specific, measureable KPIV’s/KQC’s from the Customer?

0
#106695

miranax
Participant

Absolutely.  First and foremost the customer’s voice must be heard.  THEN the Attribute GR&R applied.

0
#106696

Heebeegeebee BB
Participant

GR&R will tell you how capable your machine is…period.
“Cleanliness” still needs to be defined, not by the BB, but by the Customer.
Again, QFD and Crticality tree

0
#106697

Thai
Participant

In the end after you complete your Voice of the Customer analysis, you will probably end up with some sort of scale (i.e. a Likert scale).  You can then perform a measurement systems analysis on that system using an attribute GR&R, but in my past experience working with subjective measurements, you will probably not like the results.  My recommendation: use visual references to indicate each level of cleanliness this will improve your measurement system capability, but not likely enough to make is officially “good.”
Good luck
Kirk

0
#106721

Michael Schlueter
Participant

Tough question, Carol.
May be it’s easier to evaluate the opposite, how dirty the construction equipment is?
What is ‘dirty’ in your situation?
Is it:

non-smooth surfaces, which in fact can be smooth (i.e. shining)
sticking particles, which should not stick there
bacteria, which should not be there?
Funny as it sounds, what is the level you have to evaluate ‘clean’ or ‘dirty’? Is it macroscopic or microscopic? Where on the machine? When?
A way to measure is to create a questionnaire or checklist with a set of relevant categories, like mentioned above. Some can evaluate the whole system, others may focus on parts which your customer will inspect suspiciously. Evaluate each at 3 or 5 levels, find a total score.
You find a similar problem from nutrition industry, where people have to evaluate the quality of tea or coffee, or the taste of products. To establish quality control there they use a standardized checklist and a small team of evaluators (to capture variability). Well, they put effort in making the ‘test’ pieces look all the same, to avoid bias. Perhaps this is not necessary for your machine ;-)
If you can try capturing statements from your customers, examples by which they evaluate a machine ‘dirty’ and ‘clean’; better even ‘very dirty’ and ‘very clean’. I think you need both extremes to capture the relevant factors for a questionnaire/checklist.
Kind regards, Michael Schlueter

0
#106733

Cannizzo
Participant

Michael, that’s a good idea about the checklist.  We defined some criteria as to clean, ie, armor all the tire sidewalls.  So we might be able to do a checklist.  We need to also make sure the machines are being over-cleaned – we spend too much time cleaning, and need to reduce it without sacrificing the customer’s requirements.  On Tuesday we’re going to work on our measurement plan, and today we’re going to try to get the CCR to something measureable…that’ll probably be harder than the actual measuring!
Thanks,
Carol

0
#106735

Michael Schlueter
Participant

Carol,
Thanks for your reply. Do you like me to have a look at your results next week? If so, just post a reply to this post and isixsigma will notify me (nice feature ;-)
The QFD folks are aware of a common trap with CCR: the unspoken CCR.
E.g. nobody will ask explicitely for a car with a break. It’s just assumed a car has a breaking function. So probably nobody will raise this request. However, when the break is missing, you, the supplier, are in trouble …
How to access the unspoken, ‘self-evident’ assumption about a ‘clean’ machine?
Somehow you must think ahead, somehow you must provoke your customer (in a positve way) to say sth. like “no, THAT’s not possible: I assumed you …; without … this machine is of no use for me at all”.
Think of, e.g., presenting machines to your customer which are clean in some regions and dirty in others (e.g. show 5 differently prepared machines). He/she will certainly make a choice and you can start asking for more deeply understanding.
What is the role of ‘clean’ for your customer? Is it just esthetics? Will his/her process stop working because untidy particles cause a problem? Is it for legal reasons? Has it just always been that way? When he/she uses your machine: how ‘clean’ will it be after use? Where will it still be clean and where will it be dirty?
Questions like these should help your customer to reflect and to express his/her real (or hidden) needs for ‘not-dirty’, i.e. ”’clean”’ requirements. BTW, do you have one or more customers? Even if it’s just one customer: how many people are involved? I.e. how many different answers will you get?
A final idea: I sometimes tell a group, who got stuck with too many options: “assume you can pick only one item for some reason: which one is it?”
This helps them focussing on what is thought to be the most relevant, the most important item. Just one: the most valuable.
E.g. your customer can ask only for one region on your machine to be cleaned for a reason: which one is it?
From that point you can continue. “Ok, why is this most important for you? What’s your idea behind? And what’s about the rest?”
Good luck again, Michael

0
#106756

John J. McDonough
Participant

Carol, you have heard heebeegeebee rant on several times about QFD.  He is absolutely right!
QFD is how you get from requirements (which you have) to measurements (what you need).
Others have cautioned about traps in QFD, and they are certainly worth paying attention.  But QFD is how you get from where you are to where you need to be.
–McD

0
#106952

HECTOR
Participant

In addition to the scales to be developed and the feasibility of visual standards, it could be developed the descriptors: They will describe the parameter descriptions in every step or at least at the begining and the end of the scale (5,7 or 10 levels) to fix the parameter. It´s a very oftenly used method in several search techniques.
Good Luck

0
#187266

w. g. miller
Member

I’ll re-open this thread with a near-quote from my old company’s calculation procedure:
“Calculations shall be in sufficient detail and sufficiently referenced to allow a knowledgable individual to verify the results.”
Just what was sufficient detail and sufficient referencing was a function of which person was doing the checking, and therefore subjective quality attributes
In the past, the attitude was “give the checker what he wants” on insufficient detail/insufficient reference comments, as these comments didn’t impact the calculation conclusions.  However, once the Six Sigma group started bean counting calculation “errors”, the calculation originators started pushing back on the insufficient detail/insufficient reference comments to avoid the stigma of “error” beans.  The results were mud fights and delays in getting calculations done.
Has anybody run into a similar situation, and if so, what did they do about it?

0
Viewing 15 posts - 1 through 15 (of 15 total)

The forum ‘General’ is closed to new topics and replies.