Sample Size in Auditing - Percentage audited when performing an internal audit

[Marc]

On 8/22/02 9:57 AM, Califano, Dina wrote:

Marc,

I tried to post a question on the auditing forum, but I did not yet recieve the e-mail required to complete the registration process.

I was wondering if you could answer this question for me.

I was wondering what the industry standard is for percentage audited when performing a quality audit. The departments that I will be auditing, currently have their own staff (a no-no) performing their own audit on 2% of the work processed. The average payments processed for this department is 1900 a week. Is auditing 2% of these enough?

Well, auditors, what do you say? Can you honestly say you determine a statistically valid sample size when you audit? How do you determine what the population is when you're planning and such?

I think this has been discussed before but I couldn't find the thread in a search.

 

[D.Scott]

IMHO - I can see no problem with their own staff doing the audits as long as they are not doing their own work. After all, they understand the process better than anyone (I hope).
Next, the sample size should be based on the circumstances of the department. 38 audits would be fine if the audit results are not showing where a higher number is needed. (95% confidence would have sample size of 125 for 1900 "lot"size).
If management has determined 2% as an acceptable indicator, I wouldn't challenge it unless the results weren't giving valid results.

Just MHO

Dave

 

[gpainter]

The number 3 usually comes up when auditing a QMS. Been to different courses by different providers and talked to different auditors and 3 seems to be the number.

 

[C Emmons]

I conduct internal audits on 19 different sites - My sampling size is one days worth of documentation - that way I cover a sample of all associates in the facility as well. I work for a transportation company - So for instance - we require our drivers to sign the bills of lading in a specific format - when I am auditing a site - I audit one days worth of bills of lading - that gives me a sampling of all drivers...Dont know if it would work for your industry.

 

[Rick Goodson]

Regardless of whether you are doing internal auditing or third party auditing, there isn't enough time to do statistically valid sampling. As Mr. Painter puts it so well, three seems to be the magic number.

Actually you can draw some valid conclusions from the small sample sizes. Intuitively if a process is running well there should be very few errors. If you draw small samples from a process that has a very small error rate, the probability of finding a non-conformance is extremely small. If you find non-conformances the process probably has a large error rate. Unfortunately, not finding errors does not prove the process is in good control.

 

[Mike S.]

D.Scott said:

38 audits would be fine if the audit results are not showing where a higher number is needed. (95% confidence would have sample size of 125 for 1900 "lot"size).
Dave

Dave,

Tis' been awhile since I've went thru my statistics text for a refresher, but I don't see how you can make the above statement about a sample size of 125 for a 1900 lot size giving a 95% confidence interval if you don't know the std. dev. of the process and what degree of sampling error is acceptable. Help. What have I forgotten????

 

[JCVP1969]

I know it is not exactly the answer you are looking for but there is a random sampling statistical quality control standard that may offer some guidance!

It is not really auditing but at least something to base it on that is useful when someone queries your decision.

Anyway here is the link (give me a shout if it does not work!)https://www.ili.co.uk/cgi-bin/main_frames?DATA=49B981BE:

 

[RJFisher]

Our CB always chooses 3 samples of whatever she's auditing, so I just follow suit with our internal auditing...

 

[AndyN]

If your internal audit process replicates the CB process then go ahead with that sampling. I'd suggest that if you want to get real value from your audits then you will take whatever sample shows the process is under control. It may be one sample or all of what was 'done' in July.

It depends on the objective(s) of the audit, not just some number plucked out of the air or an emulation of another auditor who was doing an entirely different type of audit!