The first time I noticed an airline using 'advanced' statistics to sell tickets was prior to a work trip in 2012. I was due to fly to Tennessee with one of my company's VPs. Smart guy, this VP was - he broke a company record by almost a full week when he booked the flights days in advance of the departure.
His foresight gave me a little extra time to study the flight details. As I looked around on the Good Old Interwebs, I noticed that the flight boasted an 'on-time performance' of 95% (1).
This seemed pretty good. Ninety-five! We would have plenty of time to wind down after landing. But was this statistic relevant?
I suppose the answer depends. The key is slack. If a given flight is packed into a tight schedule, the on-time performance of earlier flight is probably more important. If the earlier flight suddenly starts landing late, the 'on-time performance' of the next flight will quickly decline.
But if there is slack built into the schedule, other flights landing late isn't automatically a problem. Up to a point, the next flight will still be able to take off on time. In this case, the 'on-time performance' won't decline as quickly if the preceding flight suddenly starts experiencing regular problems.
A better analytical tool might allow a buyer to use the landing time to incorporate a given flight's scheduling slack. The airline could then adjust the 'on-time performance' measure to help the traveler understand the history of the route in the context of the traveler's schedule. If a flight is always late to arrive but lands within an hour of the scheduled time with 100% historical success, I might book the flight anyway if I can afford the extra hour's delay.
I use a version of this thinking to travel to and from New York. When going down, I will try to ride a cheap bus to New Haven and switch to the local train into Grand Central. It does not matter at all if the bus is late because there is no train to 'catch' - I just get on the next available one using a ticket that is valid for the entire day.
On the return trip, I never stop at New Haven. The risk of missing the bus is too great if the local train gets delayed. Since there is usually only one bus available, missing the bus means spending the night in New Haven. No thanks! For the return trip, I always ride the train the entire way from New York to Boston.
Footnotes / imagined complaints
1. Don't like it? Sue me.
I'm making up this number. Like I could remember this years later, right? With all the beers I drink, it's a miracle I remember anything.
I'm just here to defend the formulas, not the numbers.
