Pleonasm said:
Doesn't really make sense to me from a business standpoint
Of course it makes business sense to evaluate failure rates, and in turn reliability of the product sent to the general market.
One wants a product that performs its duty in most cases.
One does not want a sloppy product, as sales will drop due to underperformance.
But one does not want to make a product that is overly engineered, as that adds to the product cost and sales drop due to it becoming overly expensive.
Its a sort of a balance of production costs, warranty costs from failure returns, to end of life of the product, so that a business does not put itself "out of business" from not paying attention of how well, or not how well, its product is performing.
Except for products that are expected to have to perform all the time, such as for military, health industry, aerospace, or for other safety reasons ( bridges, buildings for example ), in which case the infant mortality is weeded out by redundancy or by other ways.
The manufacturing process for the final product is a long chain of events from obtaining the raw materials and purification, to manufacture of small parts from the raw materials, design of a working product, and assembly of the final product.
Design can have weak points, such as the capacitors mentioned above.
Raw material purity can vary - if 97 % pure copper ingot is required and your supplier provides 97.5% you may be ahead of the game, but if your pulling copper wire, how do you actually know that each meter is 97.5% and not 97.9% in some places and 96.8% in others.
Manufacturing and assembly has tolerances for the small parts as well as for the final product.
Everything is a " law of averages", statistical sampling, with a good deal of hope, and mathematical ensurance that a minimum of product, a value assigned by you, do not turn out to be lemons.
this might be a good read.
http://www.gatewaycoalition.org/files/Enggstats/htmls/Ch4.pdf