Hacker News Reader

Comment by rootusrootus | Hacker News Reader

There is a pretty big difference between 1-2 a year and thousands. Seems like it would be trivial to use an algorithm to flag the users that are, say, two or more standard deviations from normal, and pay a little more attention to their circumstances.

View on HN

You don't even need fancy algorithms. Seems like they they could just look at the current average number of claims per customer per year, there's no way it's higher then say 5. And then just flag everything after that.

Not saying they deny the claims after that, just flag them to be looked into.

You'd want to flag the 95 or 99 percentile instead of the top 50 percent.

Not exactly meaningfully harder for a computer to do one than the other.

Probably the thing to do is work backwards from the resources that are available to investigate them. Then just skim off that many outliers. Do it until it's not saving the company more money than it takes to pay those people.

Of course, that implies that Apple has anyone to look at them, which it may not, and is probably the first problem...

Heh, looking at something a couple std devs from the mean isn't fancy; it's a pretty standard way of looking for outliers in any process. Besides, flagging everyone over the average would mean flagging roughly half of all return claims. That would require legions of people to look at, and would provide a really bad customer experience.

Reply on news.ycombinator.com