< Back
SWAPA Provides In-Depth Look at Winter Meltdown

SWAPA Provides In-Depth Look at Winter Meltdown

January 06, 2023
Dissecting the Flightmare with Key Data Points

The Pilots of Southwest deserve an explanation of the whats and whys regarding how this latest and most tragic meltdown took place. Though Flight Operations has repeatedly stated that SWAPA has been informed and consulted through this meltdown, the truth is SWAPA was only officially briefed about the plans for winter operations once, on December 20, prior to any decisions being finalized about cancellations. There have been no other formal communication or effort to collaborate on supporting the operation during the system meltdown or subsequent recovery plans. The timeline and our assessment here and in this video are based on watching the pertinent data points, answering member inquiries, and reading the same Company comm that went to the Pilots.

The bottom line, however, is that there are systemic issues within our operation, many of which have been pointed out repeatedly and solutions offered. Some solutions require Leadership and others require technology. Multiple levels of failure resulted in a near catastrophic outcome. Perhaps Pilots are the only ones taught Risk Resource Management as part of their leadership training at this Company.
Tuesday, December 20: Before the Storm 
SRC attended our monthly meeting with Crew Planning and Crew Scheduling, where we were advised MDW might need a full closure and DEN would be thinned out significantly to allow ground staff enough time to turn planes while working in sub-freezing weather. We were reassured that they were ahead of the weather event and would have enough time to process cancellations and mitigate irregular operations.  

Key Point #1: Dispatch’s Baker program does not account for crew (much)
Dispatch processes delays and cancellations through the Baker, a system which optimizes changes to aircraft and passenger routings. However, it does not account much (or at all) for crew requirements. Once the Baker solution is completed, those flight changes are handed over to Crew and Inflight Scheduling who then process the cancellations via a separate optimizer for Crews — SkySolver. Pilot and flight attendant solutions are run independently of each other as work rules governing each are different.

Key Point #2: SkySolver can only handle 200-300 issues at a time
SkySolver cannot process a high level of cancellations all at once — it can do anywhere from 200-300 at a time. Large cancellation packages more than 24 hours out are manageable though. Given the extra time, multiple runs of SkySolver solutions can address most issues, though that process of multiple iterations can cause some odd or inefficient outcomes. So, the initial plan to cancel a large number of flights seemed reasonable, as the early decision to cancel would leave enough time to cover everything. To be clear though, we (and the Company) have known for a long time that there is a need for internal SkySolver solution improvement as well as more capability for processing larger batches of cancellations and misconnects.
Wednesday, December 21: The Company’s Preparation
SWA sent an update to all Pilots that there would be no full base closure based on the forecast, but the plan was still to reduce DEN flights by processing cancellations early Wednesday for Thursday. MDW would also have a cancellation package implemented late Wednesday for Thursday afternoon through late Friday. 
 
Key Point #3: The initial Denver cancellation package was too optimistic
The cancellation package turned out to not be as large as anticipated when briefed the day prior. As far as we know, SkySolver was not overwhelmed at this point, since the cancellations were more than 24 hours out, which again allowed enough time to process multiple, smaller solution runs for pairing reassignments. Wednesday night, however, really did turn into a freezing whiteout in Denver (heavy snow, visibility one-quarter of a mile with winds gusting from the northwest up to 32 knots and a temperature of negative 8 degrees Fahrenheit) during the late evening, and flights were significantly delayed getting in and out. This placed a back-breaking strain on Ground Operations and also caused many legality and misconnect issues for crews. 

Thursday, December 22: The Operation Convulses
As we monitored cancellations, we began to see an uptick of closer-in cancellations, indicating the network was struggling to recover from the events of the night before.  

Key Point #4: The weather turned out worse
It started to become apparent that the NOC and Ground Ops leadership had overestimated the ability of DEN and now MDW and other Midwest stations’ ability to operate. Even with limited exposure to the cold, several rampers who did come to work suffered frostbite attempting to work in these conditions. 
 
Another layer of concern is when cancellations hit the Scheduling/SkySolver system later in the evening, Scheduling is already busy with flight delays and Pilot illegalities/rest issues going into the next day.

Friday, December 23: Baker and SkySolver Eat Each Other
As mentioned previously, Dispatch’s Baker passenger and flight optimizer has minimal programming pertaining to crew, and you end up with a recovery plan that looks pretty but is doomed because it isn’t aligned with crew availability. See the orange and red upticks each evening of the 22nd and 23rd as well as throughout the 24th and 25th. 
Key Point #5: Short-notice cancellations kill SkySolver solutions
When Dispatch issued more and more cancellation packages and they became much closer in, and even after departure time over the weekend, those short-notice cancellations hampered SkySolver’s ability to generate useable solutions, which necessitated more manual reroutes, causing further strain on Scheduling. Short-notice cancellations hamstring both SkySolver and schedulers as Pilots have less duty time, less remaining flights to reconnect Pilots to uncovered flying, and less time to crunch a solution. Then another cancellation package hits and unravels the prior solutions. This is another area where SRC’s recommendations, such as using Open Time as well as Last Resort Flying (LRF – outstation Open Time) might help when you just can’t cover an outstation flight except by cascading reroutes that cause so many follow-on consequences (or just a cancellation), especially when another cancellation package hits and strands crews who were part of a prior solution.
A memo then went out late on the 23rd from Flight Ops leadership with details about numerous issues systemwide impacting operation besides DEN and MDW, as well as calling out sick fatigue and Open Time participation as culprits to their struggles. SWAPA expected a drop in Open Time participation, but even with a reduction in bids, there were still Pilots who were not awarded Open Time who submitted bids because Scheduling assigned reserves instead. Additional Pilots in the system leading into the weekend could have helped mitigate cancellations with more Pilots available and online. Of course, that’s under normal circumstances. 

Key Point #6: Incentives work best before there are problems
Late on Friday, when Flight Ops leadership put out that Scheduling would be awarding premium first before reserves, it was too little too late for this solution. Automatic premium needs to be offered early to have any significant impact.

Saturday, December 24 – Sunday, December 25: The Weekend Operation Continues to Disintegrate
Chaos ensued as Dispatch processed additional waves of cancellations. While possibly warranted, Crew Scheduling had become so overwhelmed that they were telling Dispatch they couldn’t handle any further disruptions. Crew Scheduling can’t track crews with Pilots and flight attendants stuck at airports without hotels, and they can’t recover crews into actual flying when they are spread across a thin system with depleted levels of flying.

Key Point #7: CHATNOC cannot handle large, short-notice reassignments either
At this level of close-in reassignment, CHATNOC (the hotel desk) also cannot keep up with the ever-increasing number and shorter-notice hotel requirements. Crews were unable to get hotel assignments in a timely manner, particularly those near contractual and FAR duty and rest limits. Many crews wound up with no hotels, whereas others had to have release and report times adjusted for proper rest. All of these required additional reassignments and also uncovered flying. SRC estimates close to 1,000 Pilots executed duty days greater than 15 hours over the course of this event, with many put on continuous duty overnight awaiting a schedule change. Also concerning are the more than 350 instances of Pilots being assigned pairings that terminated somewhere other than their domicile, which was likely correlated to displaced crews that Scheduling was attempting to track. These are the final, ugly outcomes of the breakdown between Dispatch and Scheduling and the long-time and well-known limitations of SkySolver.
Key Point #8: Breakdown of communication between CSS and OpsSuite
Pilots started reporting that crew lists provided by ops agents from OpsSuite contained wildly inaccurate information. Flight attendants shown as working the flight on the crew list were different than the reality as displayed in flight attendant CWA/CSS. This was despite revisions to the actual working crew were made hours and hours in advance. This caused needless confusion and delay trying to reconcile the actual crew lists. The legality and misconnect issues caused by these delays — not to mention the (attempted) call volume added further workload to Scheduling.
 
Key Point #9: Losing the bubble – stranded passengers watch ferry flights depart
SWA also began using position ferries to re-align crew and aircraft positioning, which likely exacerbated passenger angst given how many of these position ferry routes had flights cancelled between those same stations. It’s unfortunate for our customers that part of the solution to stabilize the operation included flying more than 500 empty flights.
Monday, December 26: Post-Christmas Hangover
SWA announced their decision to thin out the network through Thursday, the 29th, to get crew back on track. Most of these massive cancellations were thus cited as “crew” since SWA decided they needed to cut more than half of each day’s schedule to get everyone back in position. More likely, it was due to Dispatch finally trying to get aircraft where they needed to be to align with the crew network. SWAPA wants to ensure that Company leadership doesn’t misinterpret crew cancellations as a lack of crew as that certainly was not the case. It’s unclear if there were any system glitches with SkySolver that created this need to cancel so many days out. Seeing how the system operated well on the reduced subsequent days, this seems very unlikely.

Key Point #10: “Cancel anything that is un-crewed”
A big question is why cut with an axe and just “cancel anything that is un-crewed” rather than force Dispatch to be crew-strategic in which flights were cancelled? At the surface this decision likely misconnected even more crew from airplanes as it blew up the line of flying for everyone who was still on that flight and aircraft flow. Now even more crew were having to deadhead (probably later than originally scheduled) which also takes time to get somewhere else just to finally be able to operate, if at all. At least on the Pilot side, there were very few of these cancelled flights that didn’t have both a CA and FO assigned at the time of cancellation. Without flight attendant data, it’s hard to further validate the cancellations for "Crew." We do note that there were multiple stories involving the inability to simply promote a deadheader to work a flight that likely impacted many flights that would have otherwise not needed to be cancelled.
Tuesday, December 27: The Ugly Truth About Manual Processes
COO Andrew Watterson put out a message about Crew Scheduling’s issues with having to manually route Pilots given reported issues with SkySolver’s re-optimizations following the multiple waves of cancellations, causing pairings to require changes multiple times. This is a significant admission. What will be done within the next six to 12 months to address this?
As can be expected, the number of Pilots not operating flights spiked given how much was cancelled versus how many Pilots were scheduled to work, ranging between 1,000 to more than 2,500 deadheads each day. Even before the post-Christmas cutback, we had already exceeded the number of Pilots actually out on pairings who were routed into unproductive deadhead-only and LIMO duty periods.

Key Point #11: Pilots online went unused
While it’s a necessary function to reconnect Pilots to flying, there were many cancellation and reroute decisions that created excess deadheading (and hotel sits) beyond what was required and blocked seats from passenger recoveries. While this may be a symptom of the meltdown, it does refute any notion that there weren’t enough Pilots available. They just couldn’t be used efficiently due to many of the reasons already discussed.
Conclusion
We haven’t even touched on other process and technology issues that you all saw like the inability to reconcile discrepancies in passenger counts, the inability to rebook passengers standing at the gate, and CSAs unable to contact to their own SOS help desk to find solutions. Those are deep dives that need to be done as well.

SRC has warned that other departments treat both Crew Planning and Crew Scheduling as an afterthought, sending both flight schedules and reassignments not optimized for crew on short timelines. This includes Network Planning’s original schedules used to build lines as well as Dispatch’s day-of changes for reassignments. Unfortunately, they simply accept their fate in the name of Warrior Spirit even though it continues to degrade both Pilot quality of life and the airline’s ability to handle disruptions. It’s disturbing that the Company built the NOC and pulled Scheduling into it but they still aren’t considered an equal partner — they still work in silos and they still underestimate the importance of the crew solution in making the Baker’s aircraft and passenger recovery solution work.

SWAPA and SRC also have expressed concerns for years about the limitations with SkySolver and the lack of interconnectivity between programs, as well as the many areas SWA could create or improve processes to ensure a more stable operation, most of which have gone ignored. Tellingly, as of writing this, SWAPA has yet to be invited to discuss the "postmortem" or talk through solutions. If leadership continues to turn a blind eye to investing in our frontline or include internal customers in future plans, SWAPA will be forced to continue demanding change in the C2020 rewrite.