Boy we've gotten on a sidetrack with this topic.
I think your on the right track it seems that the average return rate of consumer electronics (during the return period) is around 15-18% of which only 5% are found to be actually defective, so quite a small % of DOA or early real faults.
Assuming your numbers are correct, that would point to a roughly 0.75-1% infant failure rate among consumer electronics in general, and I really wouldn't be surprised if it was even lower for big-ticket items like smartphones or computers that probably see more thorough factory testing.
The graph that
east85 posted doesn't break out the first 30 days or DOA, but since the failure rate across the board is pretty similar between year 2 and year 3, you can be almost certain that the infant failures are greater than the year 1 failures minus the year 2 ones (since if it doesn't break right away, it should otherwise be more likely to last through the first year than the second).
Backblaze has a similar study, and much more rigorous because of the controlled real-world conditions and large sample size, for failures of hard drives:
https://www.backblaze.com/blog/what-hard-drive-should-i-buy/
There's a survival rate graph partway down in which for WD in particular you can clearly see infant failures and the general failure trend. First there's a sudden survival drop of a few points (infant failures in the first month or so), then the line levels off for a couple of years, then starts dropping at an increasingly steep rate, as the drives wear out.
Bottom line being, if it survives the first 30 days or so, it's pretty likely to hold up for the next 1-3 years (depending on product type), at which point failure becomes increasingly more likely.
And coming back to Apple, that graph shows very clearly that Apple has extremely good infant failure rates--they ship very few products that are DOA or fail quickly as evidenced by the ~2% first-year failure rate versus every other company in the 8-10% range. Logically, that tells you that Apple's infant failure rate is extremely low (probably well under 1%), while everybody else's is probably in the 5-8% range. That actually seems surprisingly high to me, but I see no other way to read the graph.
