Can/should SPC be applied to Computer Assembly and Software Imaging?

A customer is recommending use of SPC. But I am having a hard time determining if SPC can/should be used for our situation of producing specialized computers, which consists primarily of:

• Manual mechanical assembly (i.e., w/screwdriver) of COTS components
• Loading a pre-existing software image from a server
• Running a series of diagnostic go/no-go functional Q/C tests on the computers

I don't see how we can apply SPC to this situation where there is no machining, soldering, etc. Further, it is a HMLV (high-mix low -volume) environment, where jobs are often for qtys of 20 or less per order.

Where needed, torque drivers are used, and those are calibrated on a regular basis. ESD items are also calibrated on a regular basis.

I just don't see what to measure for SPC that could be useful.

The customer may be working from a default checklist obtained from a source that felt that SPC was appropriate for what they did.

Has your organization had any "quality escapes" with this customer that spurred them to make suggestions regarding your methods?

The SPC recommendation is a standard checklist item in their supplier relations; it was not driven by escapes.

Is any variable data collected or measured throughout the process? Maybe product measures?

Or you could use attribute charts to monitor defects found at process steps, testing, or audits.

I agree with the above, the requirement may have come from a checklist and mignt not be applicable. However, SPC can be observed from attribute data. If you have go-no-go or pass/fail results, these can be used to determine the control of the processes.

Yes, I was thinking that the only way I could do SPC was via attribute data. The data is essentially all binary (pass/fail), and since the job sizes are so small, there won't be much "statistics" in the SPC.

Ok, thanks folks - this was helpful!

It depends on how you bucketize the information. I have used attribute SPC in electronics manufacturing in your exact situation for many years. You can track the defects by order. But it may be more useful to look at the performance in larger buckets such as weekly or monthly to monitor the production process as a whole. The main benefit is knowing when to focus on the process as opposed to reacting to a specific defect as if it were a crisis. The benefits from attribute SPC are the same as with variable data SPC. It helps you identify when your process is out of control.

Tagin said:

Yes, I was thinking that the only way I could do SPC was via attribute data. The data is essentially all binary (pass/fail), and since the job sizes are so small, there won't be much "statistics" in the SPC.

Click to expand...

You have a good point. The whole point of Shewhart charts is to find a trend that indicates you have a special cause arise after you have eliminated all of the special causes! In attribute data, a trend may be a spike in rejections for that test. If you see a trend you can act on it. But, you won't see much action within small jobs. Maybe longer term shifts between lots.

But, yes, people love the SPC rubber stamp. It's not the only rubber stamp they like!

Also keep in mind, if you use a stats package such as minitab it will automatically factor in the varying lot sizes when determine the control limits.