Blog

THE STORY OF THE "GARAGE-O-MIZER"

Aug 07 2006

If you were to ask me how we write the Southwest Airlines flight schedule, my answer would be much different today than it was just a few years ago.  When I joined Southwest in 1990, we had just over 100 airplanes, a little over 1,000 flights a day, and about eight to nine weeks to analyze and re-write our airlines schedule by hand;which we did five times a year.
 

Fast forward to 1999.  By that point, we had grown to nearly 300 airplanes, over 2,400 flights a day, and we still had no more than eight to nine weeks to analyze and re-write our airline's schedule, which we still did five times a year by hand.  Clearly, the size of the job we had to do in the time allotted had grown enormously, the complexity of the job had increased, and our system had grown exponentially, but given our process, we had no more time nor available manpower to apply.  It became clear to the Company that we had to find a way to bring computer technology into the mix to help us work quicker and work better.  We had to be able to look at more scenarios and produce a more effective schedule that was more convenient for our Customers QUICK!

Unfortunately, in talking to experts in optimization theory across the industry (believe it or not, airline schedule optimization is quite the science!), we found that our network had become too dense and large to solve, given the limitations of the science and technology at the time.  The roadblock turned out to be our basic business model!  In a previous blog post, I explained the difference between our point-to-point network and a hub-and-spoke airline network (click here to read that post).  For a variety of reasons (which I won't bore you with!) hub-and-spoke networks are relatively easy to schedule.  Point-to-point networks are very hard to schedule because there are vastly more options in departure times, ways to "hook together" flights, and many other factors.  One optimization scientist said that, based on the size of our network at the time, there were 1023 (that's the number ten followed by 23 zeros!) of different possible combinations of our set of flights, departure times and "hook-ups" that could comprise a valid, flyable Southwest Airlines schedule.  Looking back at my years of writing ever-increasingly more complex schedules, I suddenly realized why my hair had gone gray.  To think, I had been blaming the "snow on my roof" on my son!
 

As we continued to research optimization shops, one of the Southwest Airlines Employees on the Schedule Optimization Team came back to the office after Christmas break and showed us an optimized schedule output that was unlike anything we'd seen before.  It still needed work, but it had conquered the major hurdle, treating every single flight's departure time and hook-up in the network as a variable.  We were amazed and VERY excited.  Which of the optimization companies had produced this, we asked?
 

None of them, he replied.  He had cracked the problem on his own and produced the solution himself, over Christmas break, on his home PC, metaphorically, in his garage.  He had done what was actually thought to be mathematically and technologically impossible at the time.  THE
''GARAGE-O-MIZER" WAS BORN!
 

The rest of the story is the stuff of Southwest legend.  We continued to refine the process and calibrate the over 2,300 parameters in the optimization process for what seemed like forever, but we began flying our first, fully optimized schedule in October 2004 (which we dubbed the "Optober" schedule).  The magnitude of change we were able to incorporate into the schedule was staggering.   

We continued to refine the optimization process, and by February 2006, we began flying our second reoptimized schedule, which, ultimately, generated millions of dollars in added revenue for the Company and our Shareholders.
 

Going forward, we plan to take maximum advantage of weather changes from winter, into summer, and back into winter.  Flights typically take less time in the warm weather months than they do in cold months, and optimization technology will allow us to take advantage of shorter flying times by putting even more flights into our schedule without investing in more aircraft.
 

We hope you like our new schedules, and we hope that you appreciate the time and effort we've spent in making our network even more convenient for you.  Now get out there; fly somewhere on Southwest;and exercise your freedom to move about our Country!

Similar Posts

Post a Comment

The content of this field is kept private and will not be shown publicly.
By submitting this form, you accept the Mollom privacy policy.

Comments

Bill,

Outstanding post. Let me be the first to say congratulations.

John

  • John Jamotta (not verified) — 08-07-2006 at 02:23 PM

Bill- great story! I have noticed how much more flexible the flight schedules have become since my days with Southwest. Working in corporate travel, I see the ebb and flow in flight demand based on the day of the week and your system will help you deploy aircraft when and where needed.

  • Brandon Dotson (not verified) — 08-07-2006 at 03:18 PM

OK, so next time I have to do an aircraft swap, I should blame it on computers, right??

jk!! At least then I get time for lunch!!!!!

It's amazing and mind boggling just to think about such a large operation working together and...well...work!!

Fascinating post! Can you point to an example flight that exists or a market that is better served as a result of this optimization? I am curious to see the fruits of this increased efficiency.

So did that employee on the Schedule Optimization Team receive a nice big bonus that year, or at least an Employee of the Quarter award? :)

  • Dave (not verified) — 08-07-2006 at 07:00 PM

Another great post Bill!

It is truly amazing what CAN be done when people simply ignore those who are sure it CANNOT be done.

Someone should tell the story of our SWIFT system too ... another key to our success.

  • Bob Hurst (43027) (not verified) — 08-07-2006 at 07:38 PM

The October 2004 changes were quite noticeable in the West Coast markets. Flights that had been uniformly spaced throughout the day became much better clustered around peak hours. I imagine that this was accomplished by linking these flights to long-haul east-west flights rather than simply bouncing aircraft up and down the West Coast.

Because Southwest's "pay to fly earlier" standby policy allows it to charge more for peak hour flights, this West Coast move alone must have generated a significant revenue boost. Congratulations on pulling it off!

Are your people going to publish any technical papers on your scheduling algorithms?

  • nsx (not verified) — 08-07-2006 at 09:00 PM

Bill - great post! Keep up the great work!
James Malone
MDW FA

  • James Malone (not verified) — 08-08-2006 at 01:18 AM

I don't have a web site but I am a "flying Grandma". Go to Phoenix to see the kids and love SWA. The only comment I have is, " What happened to the sense of humor the cabin attendents had? I've had some non-pluss type folks on my flights. Are you considering robots for cabin attendents and this is the training?
Bring back the humans with humor and caring, they make the flight more fun.

  • joan miller (not verified) — 08-08-2006 at 02:54 AM

Wonderful story, Bill. Another great example at SWA of individuals going over and above what is thought to be their limits to get the job done. The Warrior Spirit lives!
How lucky are we to have such employees working for us?
Thanks to you all for doing a trememdous job!

Luv ya/mean it!

  • Judy Ballard (not verified) — 08-08-2006 at 11:43 AM

Post new comment

The content of this field is kept private and will not be shown publicly.
By submitting this form, you accept the Mollom privacy policy.