Flightcaster is all about the probability of being delayed though. The big customer is business users who will rebook on a flight that has less of a chance of delay. Airlines don't want that.

Also, what's their rate on false positives?

False positives are a killer to the airlines - they don't want to risk losing customers to another flight/airline when there will actually be no delay.

To customers, false positives are likely to be a minor issue though, probably the number of true positives will outweigh occasional false positives.

But without a false positive rate near 0, I can't imagine the airlines themselves being too interested in flightcaster's predictions.

FlightCaster lies in the predictive equivalent of the 'uncanny valley'. The problem with this kind of service is that while even 90% accuracy would be a great technical achievement, even 99% accuracy is probably insufficient for practical use. How many people actually change their arrangements because FlightCaster 'thinks' the flight is delayed vs those who simply wait to hear the official word? My guess is, not enough to build a business on.

Yes.

There's a few metrics the airlines care about, the most important is leaving the gate on time. But once you're on the plane they can make you sit there for however long it is to sit on the tarmac.

When I was flying constantly for a month, almost all my flights were on time. FlightCaster did predict 2 flights that were to be delayed despite the planes pushing off on time. We just sat on the tarmac, upwards of an hour, before finally taking off.

By the same token, FlightCaster didn't predict the three other flights I was on that was delayed.

Either way, the use case is kind of limited. Even if the prediction is true and given enough warning ahead of time I have no means, and in most cases no need, to switch flights. For business travelers that have the expense accounts to make this work, this is a great tool.