Wednesday, October 9, 2019

Mistakes were made: How I failed to understand the basics

After getting the call that US Airways was withdrawing from Colorado Springs, I went into an immediate reaction to help prepare my leadership with the best possible explanation for US Airways' decision. I started to pull what data I knew. It was inevitable the press would be calling the following morning. Following my (flawed) analysis, we took the stance that the route failure was likely due to the cost inefficiencies of the 50 seater aircraft in the market. If US Airways had operated the right aircraft within the market, the route would have continued to be successful. (Press coverage)

Like most US airports at the time, we heard the 50 seaters were dead men walking and airlines upgauging to drive unit costs lower was the way of the future. Based on my analysis, our revenue and load factor performance were on par with other US Airways flights. It was only logical to believe the route failure was based solely on aircraft costs.

In reality, route level trip cost modeling based on public data is incredibly complex, unlike ticket revenues prorates. Ticket revenue prorate methodology generally follows either the square root of mileage or IATA predetermined prorates. These two methodologies deliver result in very similar results. Since IATA prorates are not public, revenue modeling used by most large data suppliers, including my models, typically follow the square root of mileage methodology.

For trip cost modeling, there is no industry standard for an airline to prorate indirect costs (aircraft ownership, maintenance, headquarters overhead, etc). Each airline's finance department determines their individual methodology. Without an industry-standard cost model, any cost model based on public data pushed to the route level will have material gaps. So much so, I would argue such models should not be used when analyzing route level performance nor presenting business cases to airlines. More on that another time.

Now, it's cringe time. When joining the industry ten years ago, I believed that the airline profitability formula was simple RASM - CASM = Profitability. The term stage length adjustments or understanding the decaying relationship between RASM and distance never crossed my mind. In my mind, all RASM was equal. No adjustments needed.

After US Airways informed us they were leaving, I produced a chart similar to the below. The below chart attempts to show COS-PHX RASM was significantly above US Airways mainline CASM. Therefore, we assumed if US Airways would just operate mainline aircraft, COS would be profitable, hence the statement "... the carrier could be losing money in the Springs because it was flying 50-seat regional jets on the route that are more expensive to operate than larger aircraft." If you didn't realize it from the graph, this is wrong. 

Further analysis backed up my hypothesis. According to DOT data, COS station outperformed the US Airways network across three core metrics we were tracking: RASM, load factor, and passengers flying nonstop on the route. Segment fares were soft but with the threeway battle in Denver, it was not surprising.

If you are new analyzing airlines, the presentations above likely did not throw red flags. At best, the RASM chart shows the bell shape curve that skews right towards higher performance. So, what is wrong? Well, it's all in the curves.

One of the first steps airport professionals do when analyzing route performance is graph a scatter plot of RASM vs distance. US Airways' COS-PHX route was right in line with the RASM curve for the US Airways' network (note: different data providers may show RASM differences). RASM curves demonstrate what an airline typically demands it's unit revenue performance over a unit of distance to be.

US Airways' 2008 RASM curve looks similar to a typical RASM curve. The shorter the stage length, the higher the RASM. Because of this, taking a simple market-level RASM and measuring to see if it is above or below the system CASM really was meaningless. Further, the bell-shaped curve that I produced for my leadership was simply a system representation of US Airways' network build rather than performance. 

Digging deeper into the performance of COS-PHX, on the surface, the route may not have caused significant concerns. The 2008 RASM appeared to be inline with system averages. However, it is important to trend the route performance over time and consider the entire environment in which the route was operating. However, it is important to note, while we do not know exactly what the route's CASM was, generally the smaller the aircraft the higher the CASM. While Colorado Springs was operating in line with system stage length RASM, however, it was produced on some of US Airways' highest CASM producing aircraft. 

Taking a look at the route's history, in the mid-2000s, the route lagged significantly in load factor (65%) with a $122 average segment fare. The US Airways' Revenue Management team clearly was not happy with this performance. Segment fares appear to be sacrificed drive up leg load factors throughout 2006 - 2007. At the same time, the Network Planning team decreased the gauge of the aircraft. All of these actions did increase load factor, but it did not actually stimulate additional passengers until 2007 when the segment fares were down 20% vs 2005. 

Further, RASM really did not respond until 2007 and stayed flat in 2008 as the team tried to raise fares back to historical levels, likely in response to significant fuel cost pressures. However, as the fares increased, demand would drop off.

In the final year of service, US Airways seemed to be heading back to decreased performance. RASM was near a five-year low, segment fares were there to match, and US was operating some of their smallest (and highest CASM) fleet within the market. At the same time US Airways was also working on upgauging and removing smaller aircraft markets from its Phoenix network. In 2008, US Airways discontinued EUG, CLD, ASE, and OKC, all of which were operated on aircraft similar to their COS-PHX route. 

Clearly, US Airways was having significant issues with fares and passengers in the market, but why? In 2006, Southwest entered Denver with an aggressive growth and fare strategy that pressured the entire Colorado to Phoenix market. Before Southwest's arrival into Denver, Denver to Phoenix fares averaged $120, however, one year later, the fares were averaging below $80. Colorado Springs fares did follow with a premium. 

US Airways did not only see pressure within the local Phoenix market but their California via Phoenix, which was the top flow markets, also saw pressure. As the intensity of the competition in Denver increased, fares Denver to California began to decline. Over the course of five years, fares Denver to California were down 25%. At the same time, fares within the Colorado Springs to California markets were relatively flat, however, passengers declined as much as 20% as passengers decided to forgo Colorado Springs for Denver's fares and service offerings. 

US Airways clearly faced a no-win situation. The growing Denver competition placed significant pressures on Colorado Springs. Any attempts by the US team to increase fares resulted in losing passengers to Denver competitive carriers. The Network Planning team was already operating some their smallest Phoenix aircraft into the market, so the ability to downgauge further was off the table. Finally, other similar-sized markets were being removed from the Phoenix network. Really, US Airways had only one choice, remove Colorado Springs from their network. And that's exactly what they did. 

No comments:

Post a Comment