Statistically Signficant "Stuff" Happens

The modern version of perfection is "5 nines", or 99.999% reliability.  This is another way of saying 0.001% failure, or 1 failure in 100,000, and 10 in a million.  Pretty good if you're making a thousand or ten thousand of anything, in which case you will probably never have a failure attributable to your product.  But if you make millions of cars, you will have dozens of failures. 

General Motors sold cars with defective ignition switches that led to the deaths of many people.  General Motors sold a lot of cars with many defective components that led the deaths of a lot other people.  Every car company sold many cars with many defective parts that led to the deaths of many people.  The difference is that GM is very big and uses so many common components among its numerous car "platforms" that their mistakes reach the level of statistical confidence. 

If big and a small manufactures make mistakes that kill people at the same rate, they will also make their families unhappy at the same rate.  But the big numbers will make the lawyers happy.   Statisticians will probably be sanguine.