# Data Points expressed in %

Six Sigma – iSixSigma Forums Old Forums General Data Points expressed in %

Viewing 15 posts - 1 through 15 (of 15 total)
• Author
Posts
• #35540

Nik
Participant

Hi,
If we measure our data points as % ( eg % of hit ratio, or % of Customer statisfaction) what is the nature of the data? Continous or attribute?
regards
Nik

0
#100270

Bob J
Participant

The data is Attribute….

0
#100323

curious
Participant

The Data technically is Attribute, but can be treated as a psuedo continous and can use tools used for continous data for analysis
hope this helps

0
#100326

R Verma
Participant

The purpose of drawing a process map is:

Uncover deficiencies.
To clarify process steps.
Get a visual picture.
All of the above.

0
#100331

mman
Participant

I agree with you,regards

0
#100334

R Verma
Participant

The purpose of drawing a  TQM Process Map is:

Uncover deficiencies.
To clarify process steps.
Get a visual picture.
All of the above.

0
#100391

Chris Seider
Participant

Nik, I’m curious the reason for the question whether a process outputing a % of x is an attribute or continuous.  I disagree with the answer you received earlier and consider it to be continuous.

0
#100397

Ken Feldman
Participant

Whether a % should be considered discrete or continuous may have more to do with the characteristic of the underlying metric rather than the fact you can carry a percentage out to many decimal places. What constitutes the numerator and denominator is what is important.  If the numerator and denominator are discrete data…sale/no sale divided by the number of sales calls…then the resulting ratio doesn’t suddenly become continuous.  If the numerator and denominator are continuous…minutes of downtime divided by total minutes..then the resulting ratio might be considered continuous.  In some cases,  percentages can be put on an I/MR chart which by nature is continuous even if the ratio is truly discrete.

0
#100404

Nik
Participant

Ok, My problem, is i am measuring a concession metric, it is
Total amount of discount given divided by no of calls received. I need to plot on a control chart to check the process control, as process is designed to have certain % of concession as acceptable limits.  Basically we try to save the sale.
rgds//Nik

0
#100406

Ken Feldman
Participant

That’s a little trickier.  I assume you have dollars over calls so your proportion represents dollars/call for some time period.  I assume this ratio could be greater than 1.  I don’t know what would constitute a reasonable frequency to report it out nor whether there would be any rational subgroup which could be made.  I would be inclinded to try an I-MR possibly by day to start with and see how it looks.  There might be further interest to look for Components of Variation and see if it varies by Call Center or Type of Call or Phone Associate, etc.  I also assume you won’t be putting the “acceptable limits” on the control chart.

0
#101479

Michael Schlueter
Participant

Nik,
At least % is something you should avoid when doing improvement activities.
The discussion gives no clear answer to your question; even worse, % won’t give you much insight into the process you try understanding or improving.
% is non-specific. Most of the time you can be more specific and gain resolution power.
However, % are ok to summarize progress:

summarizing is more global and less specific.
Kind regards, Michael Schlueter

0
#101485

Glenn
Participant

Mike,
Won’t the percentage give me a sigma level (convert % to sigma)?

0
#101486

Sphynxras
Member

I have dealt with the data as continuous in my most recent project…(process recovery rates, which are percentages), and had no issues treating the data as continuous.
sphynxras

0
#101490

mjones
Participant

The answer is, yes, if you are measuring level of performance, such as % defective, or % on-time delivery where 100% is your goal. If your % is based on ratios of continuous data, like % moisture or % butterfat, it has nothing to do with sigma of your process.
As implied in previous posts, you can even treat % of attribute measures as if it is continuous data — sometimes. To work, it should ‘look like’ continuous, semi-normal data, i.e., with % well-spread and not close to 1 (or 0).
Suggest: Just go do it. Treat % as continuous data in your analysis BUT know and understand that you are violating the assumptions inherent to a valid analysis. Get all the insights you can from the analysis, but use your results with great caution.

0
#101551

Jonathon L. Andell
Participant

Officially, # of discounts / # of calls is a proportion, which officially is attribute-based. Unofficially, if there are high call volumes (hundreds per time period), and reasonably high discount volumes (at least 25 per time period), the data could be handled as semi-continuous. The Statistics Police won’t drag you off in the middle of the night.
However, as others point out, you might want to consider another metric based in dollars. Maybe it’s dollars of discount per dollars of related pricing. Except for those who fret that pennies cannot be subdivided, most practitioners would consider that truly continuous. More importantly, you’d be connecting with the “real” business impact of the discounts, and you’d have a relatively easy time demonstrating cost savings as increases in net revenue.

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

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