I understand this and believe I have made that clear numerous times.

I believe I've also made clear that while I want to objectively engage in brainstorming and root cause analysis rather than engage in futile venting, I don't intend to let Nokia off the hook.

Let me run you folks through a hypothetical scenario:

Company N decides to make and market a widget... we'll call it widget7. The widget is unique, advanced and complex. Company N understands the potential widget market well, having been successful in parallel enterprises and engaged in detailed analysis of the potential market. Company N is also heavily involved in the infrastructure that will support the deployed widgets, another factor that should help the product's success.

Company N relies on other companies to make widget components, especially the doohickey, an extremely advanced part manufactured by the highly reputable Company S. Company S proves themselves adept at making the required doohickeys for widget7. There are, however, some defective doohickeys that make their way into widget7s.

Even as widget7 begins slowly infiltrating the enthusiast marketplace, Company N is already hard at work on widget8. Widget8 requires a more advanced doohickey, based on user feedback from widget7. Let's call it doohickey+. Company S demonstrates that they can handle it, assuring Company N they've addressed the earlier issues. In fact, the purchasing agreement makes Company S completely responsible for outgoing quality of the doohickey+.

Let that sink in.

Company N will still perform full testing on assembled products, of course, although using an AQL sample. As we all know by now, AQL sampling is designed to pass products, not fail them-- the goal is to sample a number that gives you confidence that IF your lot fails then you are not failing the entire lot unnecessarily. It is a gamble that almost every consumer device manufacturer makes, since it is time and cost prohibitive to fully inspect every single item in large lots. Let's say the lots are 10,000 per production run. Let's also say that sample sizes are 315 devices out of that lot, arrived at by standard statistical formulae.

Company N performs their testing and finds no doohickey+ defects during the production run. Testing is extensive and ensures that all critical functionality is covered.

After a couple of months, it becomes apparent that a tiny minority of purchased widget8s have problems with their doohickey+s. Based on the AQL sampling, Company N knows without a doubt that the defects affected no more than 5% of produced devices. They know this because, based on the sample technique, anything more than 5% would have caused defective doohickey+s to appear in the tested devices, enough to fail the lot and drive rework. 5% of 10,000 is 500. That seems like a lot to the purchasers, who congregate in a forum to commiserate, but of course quality engineers know that it is not. Still, the following formula applies:

Perception = Reality.

Unknown to those angry purchasers, even as defective doohickey+s showed up in the wild, they also appeared in internally-acquired devices. Company N passes this along to Company S, who was responsible for letting the defective doohickey+s out of their plant. Corrective and preventive measures are taken.

Company N's repair systems *should* be equipped to handle the ~500 devices in the wild, but for inexplicable reasons they are not. What should have been a small, easily-contained situation festers until it appears larger than it really is due to the justified complaints of a vocal minority. The situation escalates-


-and that's where I have to end our fairy tale. The real ending has yet to be written.