In Manhattan, 100% of deliveries from their in-house delivery service to my building (any tenant) are returned as “undeliverable”, while 100% of deliveries from USPS and UPS are fine.
It’s a 24/7 multiple doorman building with a famous address (even {address}.com works) and the number printed in 10 foot letters.
There are zero moments in years that a package could not be delivered, yet my account shows 50 fails of their in-house delivery, and 450 successes of UPS and USPS.
So ... riddle me this.
What does their vaunted machine learning have to say about such situations?
It’s scenarios like that that suggest the best ML still has a long way to go.