Managing flight schedules are as important for improving airline bottom lines as airport bottlenecks. Schedule padding in isolation may never achieve the desired result of improving operational robustness. Instead, it could further conceal operational inefficiencies and impose additional costs.
Flight delays and cancellations are common to most airlines worldwide. Delays can be costly and have a signiﬁcant impact on the economy, particularly when air transportation is one of the predominant modes of passenger travel. A report by the Joint Economic Committee of the US. Congress [see Schumer and Maloney, 2008] estimated that the total cost to the US economy because of ﬂight delays was close to $41 billion in 2007 – a year when airline on-time performance (OTP) was the second worst on record in the US. This includes an estimated $19 billion in operating costs to the airlines, as well as $12 billion in passenger delay costs. The report also estimated that ﬂight delays resulted in consumption of 740 million additional gallons of jet fuel, costing an additional $1.6 billion in fuel costs.
While some of the causes of delays, such as weather events, are beyond the control of the airlines, inefficient airline operations also result in signiﬁcant ﬂight delays. For example, air carrier related delays, which include activities such as aircraft maintenance, crew management, on-board cleaning, baggage handling and fueling, were the second most common problem in the US in 2014 [Kenney, 2015] and accounted for 30.2% of total delay minutes. However, it is also worth noting that not all delays are due to weather events and inefficient daily operations. Some causes of delays are also attributable to the network and schedule planning decisions of an airline. For example, while an airline develops its hub-and-spoke network, it typically does not account for the congestion externality imposed on other carriers operating out of the same hub stations. Mayer and Sinai [2003b] empirically demonstrate that the gains from hubbing activities offset the costs incurred by ﬂight delays and congestions. In a related paper, Mayer and Sinai [2003a] also hypothesise that wage cost minimisation and aircraft use maximisation result in airlines ﬂying with very tight schedules. Such objectives are typical in most airline planning systems, which are designed to achieve cost-efficient resource use.
Not all delays are due to weather events and inefficient daily operations. Some causes of delays are also attributable to the network and schedule planning decisions of an airline.
For any airline reducing ﬂight delays is important for two reasons. First, it reduces operating costs, and second, it may improve the airline’s OTP ranking. Typically, airline operations are compared on the basis of their OTP rankings, and hence airlines perceive their OTP as an important operational measure of their schedule reliability. Thus, it becomes important to plan for operational uncertainty when building the airline’s schedule. An airline schedule comprises a list of ﬂights and speciﬁes the origin, destination, scheduled departure, and arrival time of each ﬂight in the airline’s network. It also speciﬁes the sequence of ﬂights a particular aircraft must operate (aircraft rotation) on the day of operation. One of the critical components of the schedule development activity is the estimation of ﬂight block-times. A ﬂight block-time is the total elapsed time between the time an aircraft pushes back from its departure gate and arrives at its destination gate. The block-time includes several components such as taxi-out time, en route time, and taxi-in time. Each of these components is subject to different types of uncertainty (and hence delay), and the total block-time delay is the sum of all individual component delays. Similarly, a critical component of planning aircraft rotations is the determination of the turn-around-time (ground-time) between consecutive ﬂights, which also has signiﬁcant uncertainty associated with it.
From a planning standpoint, operational uncertainty that causes delays falls into two buckets: (i) the randomness in the intrinsic block-time for a scheduled ﬂight (which does not include delays caused because of any other ﬂight in the airlines network), and (ii) the propagation of this randomness through the air-travel network and infrastructure, i.e., the propagated delay [see Arikan et al., 2013]. Typically, most airlines focus on improving their OTP by adding “buffers” to the estimated block- and ground-times during the planning stages. Essentially, the hope is that an inﬂated block-time will compensate for intrinsic ﬂight delays (uncertainty) and the ground-time buffer will absorb propagated delays. However, there are several issues related to such “padding” of the schedule. First, allocating such buffers can be challenging because of the complexity involved in systematically capturing operational uncertainty [see Skaltsas, 2011]. Consequently, airline planners use ad-hoc techniques to either lower or raise block-times across the entire ﬂight network in the hope of improving OTP. Such padding can be inefficient as well as very costly [see Sohoni et al.,2011, Ball et al., 2010], because most planned resource costs, such as aircraft and crew use costs, depend on the cumulative hours in a schedule. Researchers from UC Berkeley estimated $8.3 billion as direct costs to airlines in 2007 due to increased crew, fuel, and maintenance. Nearly half the cost was due to padded schedules [see Guy, 2010]. Second, there is no clear evidence that schedule padding has resulted in signiﬁcantly improving OTP. For example, as Wong and Tsai  argue, additional ground-time buffers can potentially increase departure delays if not allocated properly. Improper allocation of ground-times may in fact increase travel times for passengers.
A more important question is whether schedule padding improves operational robustness from a passenger’s standpoint, particularly when not allocated judiciously. It may be necessary to develop, and focus on, a network-wide passenger-centric performance metric to compare airline performance, in addition to single ﬂight OTP measures [Sohoni et al., 2011]. For example, comparing airline networks based on the OTP (on-time completion probability) of multi-ﬂight passenger itineraries may be of greater value. Such a metric would also help distinguish carriers serving markets with a larger proportion of multi-ﬂight itineraries versus carriers serving markets with single ﬂight itineraries. Padding could have other problems too. Simply focusing on buffering ignores the externality imposed on other players within the air-transportation system. Because most airline ﬂight networks are tightly coupled, any delay caused within a particular airline’s system propagates across the entire transportation network and can impact the performance of the entire air-transportation infrastructure.
To improve schedule robustness it is important to recognise that delays due to different ﬂights create differential impact on the network. Understanding and accounting for these differences, identifying and monitoring “bottleneck” (ﬂights which cause the most delays) ﬂights in a network [see Arikan et al., 2013] is critical to improving the overall performance of the system – particularly in arresting propagated delay. In particular, it may be essential to allocate resources to these ﬂights differentially (not just buffers) to control their delays. Similarly, identifying bottleneck airports, where congestion causes the most delays in the system could also help improve schedule robustness.
Just as with bottleneck ﬂights, adjusting the ﬂight departure times or ground-time buffers at these bottleneck airports, and allocating additional resources during heavy traffic periods could help reduce the impact of congestion. In some cases it may be better to make the schedule sparser by removing ﬂights from congested periods.
Schedule padding in isolation may never achieve the desired result of improving operational robustness. Instead, it could further conceal operational inefficiencies and impose additional costs. Airlines need to develop additional capabilities to address the issue of planning for uncertainty, since this continues to remain a critical challenge for the industry.
Mazhar Arikan, Vinayak Deshpande, and Milind Sohoni. Building reliable air-travel infrastructure using empirical data and stochastic models of airline networks. Operations Research, 61(1):45–64, 2013.
Michael Ball, Cynthia Barnhart, Martin Dresner, Mark Hansen, Kevin Neels, Amedeo Odoni, Everett
Peterson, Lance Sherry, Antonio Trani, Bo Zou, et al. Total delay impact study. 2010. URL: http://www.nextor.org/pubs/TDI_Report_Final_11_03_10.pdf. Last accessed: June 30, 2016.
Ann Brody Guy. Flight delays cost $32.9 billion, passengers foot half the bill, October 2010. URL http://news.berkeley.edu/2010/10/18/flight_delays/. Last accessed: June 28, 2016.
Regina Kenney. Top 5 reasons for ﬂight delays, April 2015. URL http://aviationweek.com/mro/top-5-reasons-flight-delays#slide-2-field_images-1292581. Last accessed: June 28, 2016.
Chris Mayer and Todd Sinai. Why do airlines systematically schedule their ﬂights to arrive late? The Wharton School, University of Pennsylvania, 2003a.
Christopher Mayer and Todd Sinai. Network eﬀects, congestion externalities, and air traﬃc delays: Or why not all delays are evil. The American Economic Review, 93(4):1194–1215, 2003b.
CE Schumer and Carolyn B Maloney. Your ﬂight has been delayed again: ﬂight delays cost passengers, airlines, and the us economy billions. The US Senate Joint Economic Committee, 2008. URL https://maloney.house.gov/media-center/press-releases/joint-economic-committee-releases-report-detailing-over-40-billion-costs-flight-delays.
Gerasimos Skaltsas. Analysis of airline schedule padding on US domestic routes. PhD thesis, Massachusetts Institute of Technology, 2011.
Milind Sohoni, Y. C. Lee, and Diego Klabjan. Robust airline scheduling under block-time uncertainty. Transportation Science, 45(4):451–464, 2011.
Jinn-Tsai Wong and Shy-Chang Tsai. A survival model for ﬂight delay propagation. Journal of Air Transport Management, 23:5–11, 2012.
1 In a related study, Ball et al.  estimated the total cost of all U.S. air transportation delays in 2007 at $31.2 billion, including $16.7 billion in passenger delay costs and a $2.2 billion cost from loss of demand incurred by passengers who avoided air travel as a result of delays.
2 Informal conversations with airline planners suggested that airlines do not judiciously allocate block-times to scheduled ﬂights to balance costs versus operational beneﬁts.
3 Airlines share resources such as airports across networks.
4 Contrary to prevalent belief, these ﬂights may not always be early ﬂights in the day, but could be mid-day ﬂights too.