Analysis of Ten Years of Data Predicts Likelihood of On-Time Flight Performance

Analytics is central to how successful e-commerce companies analyze and make personalized recommendations for customers. At Rearden Commerce, our Deem platform relies heavily on analytics to match customers with hotels they’ll like, flights that meet their specified needs, and even relevant discounted offers for products and services in any location where they may be.

As an example, one of the most important pieces of information we account for is on-time performance for flights.  Accost a random traveler from the business rank and file and ask him how he avoids arriving at his destination late (or not at all) from a cross-country flight. Chances are you’ll hear “don’t connect through O’Hare in winter” or a similar pearl of folk traveling wisdom. But what’s Plan B? After all, winter is winter. Perhaps the alternative will be just as risky. Out of the frying pan, into the fire, as they say, or in this case, out of the ‘fridge, into the freezer.

Many common connection points are subject to weather delays. Weather is but one factor, as you can well guess. The national network of US airports is complex system that works well much of the time, but as the data show, it struggles under peak loads.

Anecdotes and conventional wisdom have their place, but the data are paramount. To demonstrate some of the analytics driving Deem, we have prepared an animated infographic depicting seasonal patterns in on-time flight performance based on the last ten years of flight data by US airport.  In this case, the data is carrier-agnostic — meant more to help us help our customers plan the best air travel routes, supported by a decade of hour-by-hour data. The video speaks for itself, but you’ll notice some interesting regional and nationwide phenomena you may not have known about.  For example:

1. Winter is indeed fraught with flight delays, but the summer vacation period is just as bad. Interestingly, July 4^th is a notable exception — averaging exceptional on-time performance. (Perhaps folks would rather watch fireworks from the ground than out an airplane window.)
2. At the end of the school year, the system sweeps from better than average on-time performance to worse starting in the southeast and sweeping toward the northwest — mapping directly to the timing that school districts set the kids free for the summer.
3. The system returns to better performance, sweeping from northwest to southeast just prior to July 4^th, then improves again across the board in late August as schools come back online.
4. Best month to fly: April. Secondarily, with some regional exceptions, September until Thanksgiving is a good time to fly.
5. Worst month to fly: December. (No explanation needed.)

SiliconANGLE ran a piece this morning on this data, citing the importance of Big Data as an area of research across the travel industry, for its vast potential in bringing benefit to travelers.  We couldn’t agree more.

Take a look at the animated infographic, and play with the playback speed settings to inspect particular regions at particular times of the year more closely.

One Response to “Analysis of Ten Years of Data Predicts Likelihood of On-Time Flight Performance”

1. Alex says:

   April 18, 2012 5:36 am at 5:36 am

   The importance of big data is obvious. Comparing data from different airlines, we can track trends and make forecasts. I really helps in this airline data from dataswell.

   Reply