4 lessons that airline revenue managers can take from Trump
US President-elect Donald Trump proves that an unconventional approach can lead to victory, and Tom Bacon ponders what this means for airline managers
Trump knows nothing about revenue management. But, of course, he knew nothing about running a political campaign and little about many key issues. But, just as he taught political experts a new winning approach, his victory has lessons for analytically driven, sophisticated airline functions like revenue management.
Here are four lessons from his unexpected victory:
1. Conventional metrics can hide unconventional trends
First, most statistical polling didn’t predict his victory – they relied on the traditional approaches to polling that had worked historically. They assumed historical turnout; they may have underweighted the rural vote. Airlines, too, need to consider new metrics in order to be victorious in the new marketplace. ‘Passenger revenue per available seat kilometre (PRASK),’ for example, has now been replaced by ‘total revenue per ASK,’ including huge new revenue streams from ancillary fees. The ancillary leaders now gain more than 40% of their total revenue from bag fees, seat assignment fees, onboard meals, priority boarding and other optional amenities. This has caused comparisons of PRASK across the industry to lose meaning. Measurement remains critical to airline management – but the wrong metrics can lead to the wrong decisions.
Measurement remains critical to airline management – but the wrong metrics can lead to the wrong decisions.
2. The missing 43%
Once again, almost half of the eligible voters in the US didn’t vote in the election – and participation varied among different voter blocks, potentially helping Trump win. Similarly, the data upon which most revenue management systems rely are based on those who buy, not those who don’t. Yet, potentially, the greatest new revenue opportunity exists in those who aren’t in the database. Low fare carriers assess ‘demand stimulation’ as a key factor for consideration of new markets and most RM systems estimate ‘spill’, demand that wasn’t accommodated on a certain flight. If airlines are turning away demand based on the selling fare and not the ‘total revenue per passenger’ (including ancillary), potentially the ‘wrong’ passengers are being turned away. Who is missing from your system?
3. Labelling people can be problematic
Trump, as a Republican, received support from long-time Democrats in the Midwest. Similarly, revenue management analysts apply historic labels to customer segments. They may segment travellers based on ‘business’ versus ‘leisure’ or how far in advance the traveller purchases his ticket. Each of these segmentation schemes, however, are admittedly crude: the business traveller could try to tie in some leisure activities during his business trip; a relatively price-insensitive passenger could book his important meeting more than 3 weeks in advance. Too much reliance on traditional airline segmentation schemes can drive suboptimal offers – for example, missing selling a business traveler a ticket for an event and missing a sell-up opportunity on a 21-day advanced purchase.
Too much reliance on traditional airline segmentation schemes can drive suboptimal offers
4. Emotional appeal can defy analytical logic
Trump’s appeal was certainly more emotional than coolly pragmatic or analytical. No one has budgeted properly for Trump’s ‘Wall’ or proven that 4% annual economic growth is realistic based on his policies. Similarly, pricing analytics often assumes the rational traveller and misses more emotional decision-making. Virgin America, for example, learned that a $19 fee for its ancillary buy-up bundle drove 30% more sales than a $20 fee. Although analytically one might assume a $1 difference wouldn’t be significant, customer behaviour proved differently.
Trump’s victory provides many lessons for functions that rely heavily on probability and statistics - including airline revenue management. The wrong metrics, potential customers missing from our databases, static labels or rigid segmentation schemes, and assumptions about rational decision-making all can contribute to sub-optimal results.