# Why Gauge R&R 10% is acceptable for variable instruments ?

V

Please let me know Why Gauge R&R 10% is acceptable for variable instruments in MSA ?

#### Ninja

##### Looking for Reality
Trusted Information Resource

I can't comment on the "why"...but sometimes the "what" can help...

What the 10%GRR {basically} means when using the TV=spec limits is that the error in your measurement is at most 10% of your spec range. It is a threshold of 'comfort' that your readings are accurate enough to tell good parts from bad.

If the spec is 17.5 - 18.5...the largest error (assuming GRR=10) is ~0.1...so the reading of 18.3 can be counted on to establish that the part is in spec.

Imagine the same thing with GRR = 30%...your reading of 18.3 might actually be from a part that is 18.6 and you've called a bad part good and shipped it.

Again, this isn't the "why"...just a way to look at the practical side...HTH

#### dgriffith

##### Quite Involved in Discussions
In a similar fashion, in calibration, there is risk that you are turning out an instrument you declared in-tolerance and good, but it is not. You want to minimize this risk, so you use a calibrator that is at least 4x better than the instrument spec.
For calibration, 4.6 times better is a risk of 2%. 10x better is very small risk.

#### Bev D

##### Heretical Statistician
Super Moderator
10% is a rule of thumb, much like a p value < .05 means that there is a statistically significant difference.

Except that it is important to remember that the % of tolerance calculation is mathematically invalid and so it's a really fuzzy rule of thumb.

Z

#### zancky

What the 10%GRR {basically} means when using the TV=spec limits is that the error in your measurement is at most 10% of your spec range. It is a threshold of 'comfort' that your readings are accurate enough to tell good parts from bad.

my modest opinion

GRR is the standard deviation of the measurement procedure error.
10%GRR means, for me, that 6 times the measurement procedure error standard deviation is 10% of TV or T according to the used formula

beacuse everything is supposed on a gaussian distribution, there is always a chance the error is above 10%. Nevertheless as long as MSA is performed on a dimension that will be controlled by X-R or X-S chart, the GRR will increase the spread of the population (R or s) so there is quite no risk in the determination of the state (good or bad)

#### Ninja

##### Looking for Reality
Trusted Information Resource
Agreeing with both Bev and zanky...

That's why I had the {basically} in there...