The airline industry is an extraordinarily difficult
one in which to operate. Competition is high, profit margins are tight,
and there are constant risks to revenue (ranging from fuel price fluctuations
to terrorism to volcano ash!) Any edge in conducting business can
add substantially to an airline business.
Optimization of certain
processes using data and statistics has become a substantial competitive
tool in the airline industry. For example, statistics have been used
to build models that can:
-
Improve the efficiencies of flight
plans, allowing the firms to cut costs (e.g. Alaska Airlines, 2013).
-
Predict the future value to the
company (customer lifetime value) of passengers. For example, Malthouse
& Blattberg (2005) used SAS® software to try predict the
20% most valuable customers, which would allow airlines to target
marketing campaigns.
-
Predict no-shows on tickets, allowing
airlines to overbook efficiently. I do note that this is not popular
with us passengers when it goes wrong, and may fall afoul of some
laws in certain countries. However, in one example, Lawrence, Hong
and Cherrier (2003) found that their predictive model of no-shows
could increase revenues in one test case by 0.4% to 3.2%.
-
Predict and improve the use of
frequent flier programs (e.g. Berengueres & Efimov, 2014; Liou
& Tzeng, 2010).
Frequent flier programs
can be substantial tools for airlines in attracting and retaining
customers. Having said this, they can be expensive, so assessing the
relative gain/loss ratios of these programs is key.
In a recent example of predicting the value
proposition of frequent flier programs, Berengueres & Efimov (2014)
worked on Etihad Airlines data from 2008 to 2012, comprising approximately
1.8 million unique passenger flight records, demographics, frequent
flier program transactions, and the like. This is already ”big
data” in that it is a dataset of substantial size, which grows
and expands rapidly as data is added to every day and every flight.
See Chapter 18 for more on the notion of big data.
In their study, Berengueres
& Efimov (2014) demonstrate statistical methods that Etihad can
use to predict key issues in the frequent flier program, such as whether
new passengers are likely to join the program soon, and whether members
are likely to upgrade to higher levels in the scheme, giving them
more privileges. Etihad incorporated these predictions into Customer
Relationship Management software, which led interactions with the
passenger.
In massive improvements
over previous predictions, the model accurately predicted 97% of new
or promoted members if three months of data was used.
The model was finally
used to create a computer app, which guides airline personnel in decisions
such as whether to grant a given passenger an upgrade.
It is such clever links
between data and real on-the-ground business decision making that
can turn statistics from a boring subject to a serious strategic edge
for organizations in the 21st century.