Roger J' Chapman

1
Real-time Airline Decision Making about
Passenger Connections
Roger J. Chapman Philip J. Smith
2
Problem

• When a large number of scheduled arrivals to an
  airlines hub are delayed due to local
  conditions, such as bad weather, a complex
  decision making situation is created with regard
  to passenger connections, leading to sub-optimal
  decisions
• Complicating Factors
• Uncertainty about arrival times
• Many arrival-departure combinations
• Multiple factors contribute to the importance of
  each arrival and departure
• Gate constraints
• Other airline decision makers are affected by any
  corrective response

3
The Passenger Ill-Will Model
(Source The Handbook of Airline Economics.)
4
Goal and Methods Used

• Goal
• After gaining a complete system perspective at a
  major airlines hub, identify causal
  relationships and redesign opportunities to
  achieve closer to optimal performance
• Research Methods
• Ethnographic Field Observations
• Process Tracing
• Structured Interviews
• Analytical Analysis

5
Observations

• The airlines decision makers studied did not
  control arrivals, but could decide whether to
  hold departures, with one composite decision per
  arrival-departure complex
• Computer support provided basic data about each
  connection leaving some data gathering and much
  aggregation to the decision maker
• The airlines hold rule involved several
  conditions, e.g. a minimum number of connecting
  passengers and a maximum estimated departure hold
  time

6
Guideline for Decision Making

• Within 30 minutes of the first departure in this
  complex generate a hard copy of the data shown in
  XXX1 including all flights arriving within 15
  minutes of complex departure time.
• For each arriving flight with four connecting
  passengers on a departing flight,
• if the departing flight is
• (not covered by a wing tip service2) and
• (not a flight that has been delayed often) and
• (not already oversold) and
• (not a high frequency service) and
• (the impact on the next use of the departing
  aircraft would be acceptable) and (hotel
  availability is not a problem if this is the
  last flight to the destination or a flight in
  the last two complexes) and (the departing
  flight would not have to be held more than 15
  minutes) and (not less than 6 (4 for last
  service) passengers are needing to connect)
  then
• hold the departing flight by including it
  in a connect message.
• Track actual arrivals for last minute changes to
  plan.
• 1 The name of the software observed
• 2 Another flight to the same destination
  scheduled shortly after this one.

7
Display Observed
8
Prescriptive versus Descriptive Decision Making

• The rule was bent for extreme single values
• The rule treated each arrival-departure
  combination separately, despite interactions, to
  simplify decision making
• The rule was replaced by simple heuristics when
  there were many delayed arrivals, due to workload
  demands and unreliable ETAs, leading to many
  sub-optimal (airline rule breaking) decisions

9
Redesign Ideas

• Better support for the airlines current hold
  rule can be achieved by
• An interface that clearly shows connections that
  meet all conditions for holding
• Supporting the efficient identification of cases
  for override consideration
• Making decision communication simple
• Including a decision critiquing option
• Providing timely feedback on the effect of
  decisions to support appropriate behavior
  adaptation

10
The All Connections Perspective
A blue background means lt 15 mins estimated time
to connect
A possible case for rule overriding at a color
border
A green background means gt 15 mins estimated
departure hold
The number in a cell is the number of
connecting passengers
Red means all holding conditions are met
11
An Analog Version of the Complex Perspective
12
Considering the Interactions between Connecting
Pairs

• Greater performance would be possible if
  interactions were considered. This might
  involve
• Developing a multi-attribute utility function for
  airline acceptable departure delay times
• An interface showing interactions between
  flights, and the utility function

13
The Single Departure Perspective
A preprogrammed airline utility function
The last arrival to delay the departure for
An individual arrival in the form of a button to
allow access to more information
This would be a widening band (representing a
confidence interval) if predictive accuracy were
included
14
Conclusions

• The guideline given to the decision maker is only
  achievable under normal operations
• These decision makers make reasonable decisions
  under increased workload situations given the
  interface
• Much of the decision making process could be
  automated, but other functions of the decision
  maker require that overall situation awareness be
  maintained
• New interfaces were explored that (1) support
  more parallel perceptual processing rather than
  deliberate, sequential calculations, (2) consider
  the interactions between more than a single
  arrival connecting with a single departure