Terminology¶
While PassengerSim is designed to be flexible and support a variety of modeling approaches, it is helpful to have a common set of terms to describe the key concepts. The following definitions are used throughout the documentation and code. They may not align exactly with the definitions used in other modeling platforms or for particular airlines or supporting companies, but they are intended to provide a consistent framework for understanding the PassengerSim model and its outputs.
Service and Network Structure¶
The following terminology is used in PassengerSim when describing the air travel services that are supplied by carriers to customers, and the structure of the airline network:
carrier represents a common carrier that operates vehicles and sells seats to customers. In most applications, a carrier is an airline and the vehicles are airplanes, but it could also be a bus or train operator, or a combination/hybrid. We often will adopt the language of airlines and flights, but the concepts are intended to be more general.
place represents a location where passengers can board or deplane an aircraft (or other vehicle). A place may be an airport, a rail station, or some other similar type location. In PassengerSim, we generally assume that each place is a distinct origin or destination for customers, and that customers don’t consider multiple places as the same origin or destination. For example, if both MDW and ORD are included in the model, we assume that customers consider each airport as a distinct origin or destination, and that they don’t consider “Chicago” as a whole as their origin or destination.
leg represents a single scheduled flight operated by a carrier, with a specific origin, destination, departure time, and arrival time. We generally assume that a leg is operated by a single aircraft in a single flight operation, although the key feature is that the beginning and end of a leg are defined by the place and time that passengers can board and deplane the aircraft, rather than by the physical aircraft operation. For example, a “leg” might include a refueling stop, but this is not modeled explicitly in PassengerSim.
service represents the set of legs that are operated by a carrier between two specific places. A service may include one or more legs, but these are legs servicing the same places, not connecting. A service is defined by the carrier, the origin and destination places. For example, a carrier’s service from ORD to LAX may include three different flight legs (at different times of day).
path represents a sequence of one or more legs that a customer can take to travel from an origin to a destination. A path is defined by the sequence of legs, the departure time of the first leg, and the arrival time of the last leg. A path may include connections, or it could be a single leg without connections. Currently, PassengerSim does not support modeling code-shares, so all paths are assumed to be comprised of legs operated by a single carrier.
market represents a pair of origin and destination places that are connected by one or more paths. A market is defined by the origin and destination places, and all paths that connect those two places are considered to be in the same market. Markets are directional, so the market from place A to place B is different from the market from place B to place A.
cabin represents the cabin types on aircraft. Cabins are strictly ordered, so it is possible to decide if one cabin is better or worse than another. This ordering is also universal; we assume that every customer would order the cabins in the same way, and that every customer would always prefer a better cabin to a worse cabin, all else equal.
Fares and Pricing¶
PassengerSim offers two paradigms for modeling fares: linear pricing and matrix pricing. In linear pricing, there is a single one-dimensional ladder of fares for any given market, and the RM system can open or close each fare on the ladder. In matrix pricing, there is a two-dimensional matrix of fares. Each cell in this two dimensional matrix may be populated with a fare or not. The RM system can open or close each row in the matrix, but cannot independently open or close each cell. Within this framework, linear pricing can be thought of as a special case of matrix pricing, where there is only one column in the matrix.
The following terminology is used in PassengerSim code when referring to various aspects of fares and pricing:
booking_class represents a “row” of the matrix under matrix pricing. A booking class may have advance purchase or other restrictions tied to it at the configuration level, but this is for the convenience of the user; when we convert from configuration to simulation all advance purchase or other restrictions are pushed into the
Fareobjects. Booking classes may also have a minimum cabin type attached.brand represents a “column” of the matrix under matrix pricing. A brand may have restrictions tied to it, or ancillaries bundled with it. A brand may have a minimum or maximum booking class (e.g. “basic economy” is never offered above booking class “Y2”, or “gold elite” is never offered below booking class “Y5”)
fare is a “cell” in the matrix under matrix pricing. It is the combination of a booking_class and a brand. It may also have uniquely defined attributes from a configuration. If a user wants to define restrictions that are not homogenous by booking_class or brand, they can do so by configuring restrictions at the fare level.
fare_class is a combination of a booking class and a brand. It is used to label the specific fare that is being offered to a customer. It is a generalization of the fare: while the
fareis specific to a market and also has other attributes like prices and restrictions, thefare_classis just a label.
Prices¶
In contrast to a “fare”, which refers to a combination of price and restrictions, when we are referring to “price” we are referring to a pure currency amount, without any restrictions attached. There are various types of “price” values that are relevant to different parts of the code, and it is important to be precise about which one we are referring to in different contexts.
customer_price is an attribute of an
Offer. It is the amount of money that a customer pays to book that offer, exclusive of any ancillaries that are not bundled into the offer. This is also the carrier revenue actually realized if a customer books an offer. This value is specific to each offer, and due to continuous pricing effects may not be exactly the same as thefare_price. In the absence of continuous pricing this value should be the same as thefare_price.fare_price is an attribute of a
Fare. It is roughly equivalent to a “published” fare for a particular combination of booking class and brand, for a given market. This is the starting point for continuous pricing adjustments when generating offers.reference_price is an attribute of a
Demand. It is the pricing anchor point for customer choice processes among customers in that segment for the given market. Thereference_priceis typically the lowestfare_pricefor any fare on any carrier serving that market, regardless of whether that fare is actually offered to the customer or not. However, this is configurable, and in some cases thereference_pricemay be some other value, or some multiple of the lowestfare_price.decision_price This is an “average” price attached to a
PathClass(paths by booking class) orBucket(legs by booking class). This is the default value that is fed as the input into all RM steps that need to know the price of various booking classes that may be opened or closed. The algorithm which determines how this average is calculated and how often it is updated is configurable.marginal_rev This is the expected positive marginal revenue to the carrier from keeping a booking class open on a path or leg. For cases where expected marginal revenue is negative, this value is zero, as we assume that the booking class will definitely be closed. These values are the result of “fare adjustment” algorithms (e.g. “MR” or “KI”) when used with variations of forecasting that account for the probability of sell up (e.g. Q or conditional). This is a vector of values by timeframe, as the probability of sell up is not constant over time. If using “standard” forecasting that has independent demands, the marginal_rev can still be expressed as a vector but all the values in the vector are equal.