A technical view on personalization – Part 1 – The passenger

by Philippe R.
in Digital Technologies
on 2018-09-25

To implement hyper-personalization and seamless travel inside an aircraft, specific aspects need to be taken into account :

Identify the traveler
Know the traveler
Adapt the aircraft
Update the knowledge

Crew and passenger are traveler, but the environment to consider are different so the following talk will separate the two use cases (see Part 2 – The crew)

Identify the passenger

Unlike a car or a business office, the aircraft is a public environment where traveler go in and out. This imply that the matching of the identity to the profile is critical for both privacy and success of the product, but also the display of the matching success shall be sometimes protected from others (for example to avoid revelling a credit card number) . As highlighted by the book/movie “The Circle“, privacy is not only a matter of transparency but also respect of others and ability to decide which information are shared and which are not with the surrounding.

For now, even if there is some leading companies in the identity, management of the identity is still a combination of various domain : service identity (like google), social network identity (like Facebook), booking identify (like Amadeus), financial identity (like credit card), brand identity (like frequent flyer), physical identity (like a phone) ….
On the other side, the protection of personal data (like GDPR) limit the persistence of the full identity of a traveler and give them the right to erase it.

This mean that the system needs to provide several identification processes, correlate to an anonymous profile, and link any further activity to the profile.

As the identification of the customer can be partial (like a credit card) and a customer can own several of them, the system needs to perform regular consolidation of the anonymous profiles.

For example, you use a credit card in March (profile 1), log with your social network account in May (profile 2) and in October purchase an item with your credit card while connected, then the system can consolidate both identities (profile 1 can be attached to profile 2)

This reconciliation can be explicit (like acknowledge request) or implicit (done by the system)

As the personal profile is anonymous, the data collected also help enhance the group profiling ( like teenagers) even if the exact identity disappear from the system (by an explicit request or a simple control of the age of the information).

Know the passenger

To know the passenger is a combination of two aspects : Historical data and Behavior model
A simple example of this is the logic of your personal car : based on the average MPG and other parameters, the car will be configured by default in sport or economy mode.

In an aircraft, the knowledge of the passenger will have several constraints

The traveler is not frequently in the aircraft
The aircraft does not always have a connection
The travel goal will affect the behavior
The traveler is moving from one aircraft to another
Each aircraft does not have the same system
Each aircraft is not provided by the same company

Let walk to each of those points

Traveler is not frequently in the aircraft

Once a traveler is identified, even if traveler can be pre-categorized based on general information (like age, cultural clusters in the world …), the core of hyper-personalization is to have frequent contact with the user to adjust the behavior model.

With an average of 4h by flight and movies with an average size of 2:30h, the number of interactions during a flight can be very low.
So the knowledge of the passenger needs to be increase outside the boundary of a flight by a accumulated amount of data.

For frequent traveler, the accumulated set of information can be enough to create a reliable profile, but, for the majority of the passengers, the contact points are too scarce for the profiling.

That means that, except for some mass communication, like kids content or advertisement, the profiling needs to look for external sources of information.

One basic way is to start the experience with an optional interaction with the user that ask for general preferences (like voice training for voice recognition software), but the additional data gathered will be limited and proportional to the time that the user accept to take for this interaction.

Background enhancement are preferred based on additional profiles information that can come from identity sources (like social network) or other travel partners (like OTA – online travel aggregator).

However, even if the association is done on the background, it is important to have the traveler aware of this association so that the relation stay on a trusted win-win collaboration : data exchange against better service.

The aircraft does not always have a connection

Unlike many static systems, the aircraft will have many situations where connectivity cannot be provided during operation.

Unless there is regulation and technologies changes, based on altitude (for example 10000 ft) or physical location (like some countries), connectivity cannot be provided to the passenger.

This alternate behavior involve that many functions are designed to operate in both modes.

The strategy can be different if the behavior is mandatory (like passenger sign in) or requested by regulation (like credit card transaction).

As the aircraft is a public space with a temporary allocation for the traveler, the mechanism cannot rely on previous authentication and static data (like your computer that store your identity after a first successful login)

However, pre-caching of the information can be implemented when the connectivity is available

Passenger information on a server that will be matched later at the seat
Content for on demand
Behavior models

For other non critical functions (like voice control), the system can apply a logic of automatic activation/desactivation of the behavior

The travel goal affect the behavior

As aircraft is still an expensive transport vessel (with time organization constraints), the passenger behavior is often influenced by the goal of its travel : leisure, business, …

As mentioned in “Is the passenger coming to or with an aircraft ?“, the passenger will have several behavior profiles in regard to those goals motivated by parameters like business deadlines, paying for his own expenses, traveling alone or in groups …

The impact for the personalization is to allow, if needed, the segregation of the data collected in each “mode” and selected by an explicit traveler choice (for example, an employee can choose on his business profile to avoid any mention of his personal life like family members …)

The traveler is moving from aircraft to another

Many travels will have connecting flights to reach out to the destination which imply for the passenger to move from one aircraft to another.
Like many travel scenario (for example an hotel room), the aircraft needs then to follow the pattern : prepare, use, clean.

In term of digital, this mean than the aircraft needs to

receive the traveler information, the content (medias and application) and show an initial airline branding setup
serve the traveler while flying
transfer back to the next vessel (or to a persistent storage waiting for the next flight) the change and addition to the traveler profile and remove any trace of the traveler on the aircraft.

This high mobility of the data is not unique, but the restriction of the bandwidth and other constraints (like system acceptation) create the need to separate the management of the system into several life cycles :

the airline identity : system configuration and initial branding
the entertainment content which will be most of the time, a combination of planned updated (like monthly content or daily news) and on request.
the traveler information which have to happen every flight.

As a last constraint, the passenger can move from one seat to another even at the last minute.

The traveler information need then to be provided from a central location (like a connectivity server) to the traveler environment (for example the IFE display of the seat) after passenger identification.

Each aircraft does not have the same system

In many industry, the global system of an operating company is a patchwork of several versions of the platform.

The diversity is, in average, proportional to the pace of update of the technologies or configurations. In the avionic industry, even if the pace is higher in the cabin than in the cockpit, the evolution rate is slower than the consumer industry.

On top of that, many airlines have several providers for IFE or connectivity systems on their fleet.

The challenge is then to define operating standards that can be used to create a seamless experience.

Digital avionics services needs then to be designed to accommodate those majors variations :

Variability of the aircraft type or generation (like A320 compared to A380)
Variability of the cabin system
Variability of the consumer part of the service (mobile devices for example)

Most of the time, the variability is managed on the part of the system that have the higher pace of evolution (the mobile and cloud components of the solution) but the other parts of the system needs to auto-declare their capabilities to facilitate the adjustment of the behavior.

This adaptations are also go deep into the data flow as the system need to account for adjustment of behavior based on available features (like a satellite connection or seat control)

Each aircraft is not provided by the same company

Even if airlines are looking to create customer attachment to their brand, the traveler are often traveling in one year with several airlines.
This mean that the fragmentation of the knowledge of the passenger is spread across each airline.

Also, each airline is developing its own digital system which create also variation in the way the data is collected and stored.
To allow cross-airline sharing, the industry need to define global standards for communication.

Like the identify subject, this does not mean that all communication needs to be leverage to the shortest set of information.
Each airline (or provider) eco-system can leverage extended functions using the communication standard as the common shared knowledge of the passenger.

For example, the information of the movie genre preference can be used for initializing the next experience but the seamless experience to resume the previous movie can only be possible inside the same airline travel.
As a reminder, in digital, integration capability is often more valued than the data retention (historical knowledge).

Adapt the aircraft

Adapt to the passenger close environment

It can seems obvious to say that the goal of the personalization is to adapt the aircraft to the user, but as mentioned in “Hi, I am red“, the adaptation is not only restricted to advertisement purpose.

Also even if the adaption of the environment can be first limited to the main device (seat or PED – Personal Electronic Device), the hyper-personalization is looking to take into account all the elements of the passenger environment :

Seat light and ambient light
Window (physical or electronic)
Physical seat
IFE screen and controllers
PEDs
Wearable devices
…

In business class or specific configuration, the personalization is also extended to a group of seats (like family business class of Qatar Airways)

The passenger environment needs to be considered as a close eco-system created for the time of the flight. This eco-system can be split into 4 elements :

The public cabin part (window, ambient light, connectivity server, IFE servers …)
The crew eco-system (aircraft or mobile devices)
The private passenger aircraft space (Seat, IFE screens and controllers, …)
The private passenger personal space (cell phone, tablet, smart watch, health devices …)

As the interaction with the crew is not permanent and the public cabin space is interacting with the passenger on some global operations (like PA passenger announcement), the control of the environment is split between the private aircraft space and personal space.

The control from the personal space can be implemented based on consumer standards (like Amazon Alexa or Google assistant), but to manage aircraft specificity and more durable interactions, many airlines are developing the concept of a companion application that facilitate the interaction of the user with the system.

The control of the system is then a balance between the personal device companion and the aircraft seat controller.

As the personal device is more likely of newer generations than the seat device, the trade-off between seat and device control is based on the willing of the passenger to have his device used (capacity and battery) during the flight for a better experience, and other parameters like convenience and accessibility of the PED.

Adapt to the moving aircraft

The main reason why we take an aircraft is for its ability to move fast from one location to another. The drawback of this ability is that the aircraft physical location and situation is changing at a high pace : from ground to air, from a country to another, from an airport to another.

Each change will impact the ability to provide connectivity (like physical location and legal restrictions), the path of the data exchange (like one satellite to another) or the location of the expected source (for example from Google.fr to Google.com).

As the aircraft is shifting from one situation to another, the system can have strategies to avoid passenger impact.

For example, at arrival at the airport, the system can provide connectivity through cellular until it reaches the gate where it switches over wifi.

When the service has to be impacted, the operational procedure need to be adapted to avoid inconvenience for the passenger.

For example :

The use of a companion application imply to have the application installed on the PED while connectivity is available or to cache the application on the IFE/connectivity server
The lost of connectivity below 10000 ft needs to trigger some backup or transaction ending behavior
…

Adapt to the configuration

As each aircraft can have a different configurations (IFE, connectivity, controlled seats …), the system needs to seamlessly adapt the experience and, also use information on travel to anticipate the incoming behavior.

A direct adaptation of the experience is for example to propose light dimming during a movie projection

An anticipation adaptation can be to highlight to the user that the flight is ending in 45 mn and that his device need 40 mn of charge.

Behind those visible behaviors, the system needs also to activate/deactivate functions based on the system situation. For example, a voice recognition that use cloud services needs to be deactivated if the connectivity is not available.

One of the main system analysis is to define the value of a global configuration (like IFE + Connectivity) compared to each element independently (IFE alone or connectivity alone).

This value is, of course, affected by the dependency of the data exchanged between the systems, in respect to global or provider specific standards.

Update the knowledge

At the end of the flight, as the system will be prepare for new owners (crew and passengers), the information need to be transferred back outside the aircraft to ensure continuity of service and resilience.

Based on the configurations, this transfer can happen through the companion application and/or through the cabin system.

The information needs also to be consolidated between both components, flights and even airlines if we look at consistent profiles.

The update of the knowledge apply to 3 areas

The passenger preferences (like default meal)
The passenger history (like last movie)
The passenger profile (like interested in sport …)

Of course, if the system is smart, the 3 aspects are inter-connected, but in average :

the preferences are based on explicit declaration
the history are facts
the profile is a system guess based on the 2 previous group of information.

As highlighted in “The 3 models of data“, the goal of profiling is also to be able to limit the amount of historical data that needs to be collected (or retained) in the system to adjust the behavior.

In a constrained environment like an aircraft, where space and processing capacity is limited (when disconnected), the profiling is a good strategy of data optimization.

A combination of local cached profile and live access to the full profile is a pattern that will allow continuity of experience despite the state change of the system (flight phases, …)