Putting Ourselves Out of Business

it should be possible to make perfect products the first time so that test becomes unnecessary.

I don't know about recently, but at one time motorola came close.

Testing is always needed, in never heard of a Project/Product that was First Time Right. it will reduce the test iterations but testing will never go away.

If you think it is perfect Murphy's Law will kick you in the @s....

I worked with industry mfg. Material Handling Equipments. It was mandatory for such products to be tested before delivered. Each unit had to be certified. So no possibility of skipping the test.

Suresh Sharma.

Now, there is two kind of testing, one is testing of prototype and another is testing of finished product before it go to store/buyer. If it is another case, and from reference of product samples it seems to be, then it would be better that every product is tested, because it saves money and create good name if faulty product is not sent from factory, and could also save lives if this is vehicle or anything that can put user in danger..... Beside, it gives people something to do, specially since there are more and more robotized factories and people have less and less to do. Even if robots would make far less faulty products, it could happen that some tool become misaligned and whole series of product could then be faulty, which would not happen if every product is tested obligatory. If INTEL has tested his Pentium 90 MHz CPUs at least by sampling, than that error on 4th. decimal place below decimal point in calculations would be catched in time. I lost 6 months bashing my head to understand why my perfectly working astronomy program work on ATARI and not on PC.....

But for testing of prototype, this could be nicely done on PC, and feedback loop could be created to make product perfect. There is one branch of programming that is called Genetic Algorithms or Evolutionary programming, that can >>mutate<< product parameters and also >>crossbred<< severall solutions to same problem (if we imagine product is such solution) to >>evolute<< best solution that most closely confirn to the requirements. That way product is superoptimized after fer hundred or thousands of >>generations<< where only best solutions >>mate<< and have >>offsprings<< while worst (least fit) >>entities<< are eliminated at same rate as new offsprings are created, but there should allways be some predefined number of entities >>alive<<. There testing of >>virtual prototypes<< serve to clasify them and determine how many >>offsprings<< they would have as best become >>dominant<< to pass their good qualities to next generation. Since best virtuall prototype is choosen, if program part that perform tests is giving correct results, there is no need for making severall prototypes, and production can start at once, confident that if all is made according to engineering plans, it should also work as expected.......

That way testing by prototype can be eliminated in future definitively. It would be up to sceptics to test products in >>real life<< conditions if they would like to spend the money, but program author should guarantee accuracy of test results.....

IMHO, engineers should spend more time creating new devices then for testing, maintenance and repair of allready produced devices.........

I could make (and I plan to) such program that would be able to test any kind of device in any imaginable condition or combination of conditions, and >>evolute<< superoptimized device for that function. I shall start as soon as I earn enough money to be able to hire Physicist to help with simulations, and I can sit and program it.

Moreover, I plan to make this program available for downolad for free by anybody who would sign contract that this person would pay 20% of selling price of any and all products that would be created by program.

What do You think of this idea?

Regards from Zagreb, the capitol of Croatia, Europe!

Marijan Pollak, IT SA/SE 1st. Class, Instructor and Team Leader (retired)

As long as the product and process is directly or indirectly controlled by us flawed humans, testing (from 100% to statistical, depending on the risk) will always be required. Even if your product or service were "perfect", how would you know? Wait for your customer to tell you? Could you maintain perfection without testing?

I had the privilege of meeting Dr. Deming at a 3 day seminar (I was his bead counter), and it took me many years to realize what he meant when he said to us that "in business there is no status quo - you are either improving or dying". He also said "measure what you want to improve"!

I believe testing will always be necessary; even in prototype level. It is true that computer simulations can take us to perfection, but we should note that simulations only simulate the conditions/environments we give them, and thus they take into consideration only the conditions we know of. There might be laws (or effects, or interferences) that we might not be aware of that are in effect for our new product. It is true that inside well-established borders of physics we will mostly be secure; but there is always the chance, no matter how slight, that something that hasn't been discovered or taken into account before might be there.