Other than First Pass Yield and using Pareto chart, are there any other metrics that we can use in Test? Also, we don't get boards for test on a consistent basis. How do we measure our productivity?

Other than First Pass Yield and using Pareto chart, are there any other metrics that we can use in Test? Also, we don't get boards for test on a consistent basis. How do we measure our productivity?

Have you considered attribute-type control charts (http://www.qualityamerica.com/knowledgecente/knowctrWhen_to_Use_an_Attribute_Chart.htm)?

Stijloor.

Other than First Pass Yield and using Pareto chart, are there any other metrics that we can use in Test? Also, we don't get boards for test on a consistent basis. How do we measure our productivity?

How about tests completed in a timely manner, tests completed correctly, test backlog?

Yield is not produced by test, it is set upstream in manufacturing.

Measures of Yield often lead to pressures to improve results with unintended results. I have been involved all too often in big efforts to prove the test method is incorrect, all the while the process keeps running poor product while the focus is on the test.

Production people live and die by Yield and they are expert at gaming this measure.

I argue that Yield should be number you said were good - 2 times number scrapped/reworked at each downstream process - 5 times warranty claims (account for lost opportunity cost).

This can lead to negative numbers. This idea is a cousin of cost of poor quality and gets about as much interest/support.

How about tests completed in a timely manner, tests completed correctly, test backlog?


I was pondering this, too. But, without a constant flow of product (and if I get the picture correctly, the product flow is intermittent) you have to log your work when you do it (painful overhead), rather than look at the total for a day (not so painful - but erroneous for this case). Not sure if the effort will yield an equal benefit. :cool:

First pass yield and a pareto of errors is sufficient to introduce metrics in the test department.

Some companies monitor test times versus the standard time and others measure the time for set-up.

What is important is that the test data gets fed back to manufacturing so that the process improves.

My tuppence worth - measuring an inspection operation or department on their throughput / accepted could lead to inspectors being put under pressure to pass non-conforming or borderline product. (It certainly happened where I work where inspectors used to work for production & were measured on the same metrics as production ie getting numbers out the door).

Would it be possible to measure them on external non-conformances (where they passed failed components)? Or as DannyK says - the inspection time taken could be used, however there is still a risk the inspectors will rush work to achieve targets and non perform the inspection correctly.

I'm not saying the rest of production shouldn't be monitored by throughput, however there are risks involved with measuring inspection in the same manner.

First pass yield at any station is becoming an accepted metric. It is up to management - as always - to recognize that where the yield loss is detected is not necessarily where the defect was created. Additionally yieid is calculated as the number passed, not as the number ultimately shipped. Any rejection - rework - resubmission is counted as a failure. This allows visibility to the hidden factory. This is now a common lean and six sigma metric. We use it on every station for every product quite effectiveley; and no one blames test or inspection as it's not allowed. I would add SPC trends for each pareto item (We call it: trend - pareto - trend. clever huh?). although we trend these as defect rates and not yield...

I also like tracking output vs the standard time. This helps get a better look at productivity when volume is variable. Alternativley you can use a ratio of output per man hours actually worked but that typically requires a separate tracking system. While the first method is pretty standard for most ERP systems.

As an aside, I don't advocate not usign a metric because others might misuse it. If the metric is correct, educate the 'others'. It's not easy, but nothing worth doing is usually easy....