Desolate Distribution
Artificial intelligence can break fertile ground in airline distribution management
This introduction is a story about the delish chocolate industry. Rumor has it that Hershey suffered quite an embarrassing episode back when it was still building its brand. Namely, they forgot to account for the weather.
They had just signed a large shipment and, of course, were eager to deliver it as soon as possible. Unfortunately, the sweltering summer heat surprised them. By the time the delivery was made, the products had thoroughly melted and become useless. Not only did the company have plenty of cleaning to do for their cargo transport, but they also learned a valuable lesson about preserving your cargo during transit.
This certainly marks it as one of the most memorable funny stories from the supply chain industry.
In the airline industry, you can’t really preserve your goods for long. They are considered perishable goods because you can’t sell it once the airplane door closes [i.e. the core transportation part - other products and features you can].
How the ‘goods’ (seats and related products and attributes) are made available for sale used to be through a retail model. But then in the ‘60s, the rapid increase in the use of travel agents of course coincided with the accelerated adoption of the wholesale model driven by Global Distribution Systems (GDS), such as Amadeus, Sabre, and Travelsky.
Airline stopped focusing on selling directly to end customers; the passengers. Only the corporate travel and travel management company (TMC) segments received considerable attention.
There’s actually a lot of applied business psychology in distribution, particularly about perceived power struggles and anticipated market response (read: supplier). And I liken it to opposition behaviors or disorders, like digging in your feet.
Today, there is a myriad of indirect-GDS, intermediary-direct, and direct technology hook-ups plus a complex mix of negotiated fares, net fares, commission-based accounts, and the re-introduction of commission-based fares provided they are booked through the direct New Distribution Capability (NDC) interface.
One aspect has become overly complex:
How to track, analyze, and optimize channel costs and content. What I mean is how to control what is sold through which channel to optimize:
(1) reach at the highest yield,
(2) breadth at the lowest transaction cost, and
(3) balanced operator-supplier-buyer market power between the GDS’ and intermediaries’ preferred procurement channel.
But Artificial Intelligence can re-level the playing field with smart technology and techniques to win back customers and optimize how seats are sold and through which channels and Points of Sale (POS). This also helps manage distribution costs.
Read on to find out more about use cases for AI in airline distribution.
Business interest
The rigidness, inflexibility, and notably the cost of using GDS’ increasingly has become an irritant to airlines that wanted to be able so have their own direct sales channel, and possibly connect large agency communities to this lower-cost channel.
This changed with the advent of the Internet in the ‘90s, but retailing had been a lost skill. What followed were frantic attempts to deploy and improve online booking engines. That has not deterred the rapid growth in power of new intermediaries, the generation of online travel agents such as Booking, Expedia, Kayak, and the like.
Today, anybody can sell travel. But that’s not the issue.
With locked-in agreements and the GDS’ power of economies of market dominance owing to comparison-shopping choice (breadth and depth), and travel agents’ reluctance to change habits (even screens they use), the landscape has essentially changed less than expected.
In recent years, there has been strong airline pressure to move out of these GDS’ by establishing direct (and since 2015 NDC) connections to avoid the steep segment fees and to overcome the limited capabilities of distributing richer content, like images, reels, or other retail products and services or features like virtual and augmented reality.
The contracts with GDS’ themselves are often an obstacle and limit the airlines’ ability to innovate in the distribution landscape, other than their website.
Many airlines, such as Scandinavian Airlines (SAS) in October 2022 negotiated an overall package that would see it distribute improved content (full access including lower fares through New Distribution Capability-NDC Direct Connect technology) in parallel to its existing Amadeus GDS contract.
Lufthansa and Amadeus have announced similar ‘NDC-based GDS agreements’, following those that had already been announced with Sabre and Travelport. The arrangement was called the "NDC Smart Offer" and is available only on a bilateral basis to agencies.
Cory Garner, CEO of T2RL, in a LinkedIn post, explained that it offers the best content and does not involve the distribution cost charge that Lufthansa maintains for GDS bookings made through other agencies for other fares[i].
These changes are causing shifts in the distribution industry (the intermediaries) but there appears to be a looming integration between the GDS’ and NDC content aggregators.
It was inevitable. Most aggregators might even be acquired by the GDS’.
* * * * *
Steering demand
What the changing landscape brings to bear is that the multitude of arrangements and the complexity involved in understanding all the (transaction) costs involved as well as the impact that distribution cost charges (i.e., fees laid on top of fares sold through certain agencies) may have on demand complicates the steering of demand.
This is especially the case if there is any focus on margin and optimization beyond the initial cost recovery of channel costs today.
This complexity can of course not be handled by a human or set of business rules in these channels without prescriptive analytics in real time to provide smart automation. It requires a combination of AI.
Further, due to the increasing airline industry consolidation (despite a flurry of new startups), negotiating corporate travel agreements has become even more difficult.
For example, airline mergers can lead to a shift in the negotiating power when the travel buyer appears relatively smaller on the newly combined carrier’s list. This happened when Brussels Airlines was acquired by Lufthansa, for instance.
Also, airline mergers that cause a rationalization in networks can make the combined carrier less of a fit for the travel needs, in terms of destinations, connections, and frequencies, especially for companies with headquarters in smaller cities, like St. Louis. This happened after the AA-TWA acquisition.
* * * * *
AI-powered use cases in distribution
There is a great potential to apply AI in use cases around the optimization of distribution based on the need or expected importance of content penetration in each of the channels.
Airlines will always be encouraged to drive bookings to direct, lower-cost, channels that also enable better engagement with end customers. But doing so can undermine their negotiating power with traditional intermediary channels like the GDS or traditional agencies using the GDS’.
Further, airlines have contracts in place with GDS’ that dictate what restrictions airlines have on offering content (e.g. the lowest fares) exclusively outside the GDS.
Building in the rules and basing potential channel shift on cost/benefit and risk cases will involve calculations that AI can assist in. It will be able to generate recommendation that also reflect the risk of losing customer segments, even at the individual corporate customer level.
Given the trend to offer full and enhanced content through NDC-enabled distribution and the aggregators’ intention to gain traction in the agency community, there will be many opportunities for more decision intelligence in:
(1) impact analysis,
(2) contract negotiations,
(3) operationalized channel shift mechanisms, supported by MarTech solutions and loyalty products.
* * * * *
Therefore, the following is a list of use cases for AI in optimizing distribution:
Channel optimization / shift based on cost, retailing optimization, propensity to spend and new revenue-generating products.
Point of sale pricing based on observed willingness to pay with ingested insights.
Embedding payment services and optimizing attributes in fare products.
Improving agency sales contracts and commission levels, notably with large online travel agents (OTAs) like Expedia, Kayak, Booking.
Distributed AI and distribution penetration using NFTs in blockchain.
Text-to-image generation using AI for NFTs to create unique artifacts.
Enabling agencies to extend the benefits they enjoy to their customer base through B2B2C metaverse where they hold a space.
It is evident that the above examples of use cases represent an illustration of the possibilities.
For instance, coupling loyalty to distribution channels and sales programs, a more complex mix of optimizations could be run using deep learning. This applies well to corporate loyalty programs but also for very frequent flyers that also shop around.
This is a very promising idea of a roadmap that would elevate the skills of the teams concerned with this with the help of smart technology. This is where I proposed Hybrid Intelligence and how to best apply it (see article).
* * * * *
Artificial
There is nothing artificial about using applied science to improve sales and distribution steering. In fact, it’s become impossible for humans to make real-time decisions on complex processes like omni-channel impression (content) and transaction optimization (max rev./min cost) where AI itself is also used to feed recommender systems on what to offer to whom.
These intersections are complex. We cannot over-simplify them with concepts.
So where do we start?
We start by looking under the hood with detailed analyses of what different behaviors can be detected, even at the same individual traveler’s level. These patterns will later serve as input for machine learning.
We continue with laying out the goal by segment and identifying all the cost per booking by transaction and channel, and relating this to risk.
There are about 10 additional levels in between where combinations of AI can be built, each potentially laid out in an artificial neural network using deep learning. It’s something we can discuss if you are interested in exploring this.
These layers can be drawn out from the outside in (goal-centric), and built in steps. They are the tip of the iceberg, but, like with Hershey’s chocolate, properly executing it by adding small-step layers of complexity can ensure the inventory is not perished, does not melt under the sun, but brought to the attention of the right buyer at the right price, transacted through the lowest cost channel, consistently.
Wishing you all a wonderful day, and greetings from a Chocolat Favoris in Montréal.
Ricardo
Montreal, Tuesday, 14 March 2023
Feel free to contact me for questions, comments, or a chat:
ricardo(at)pomonaadvisors(dot)com
my general email has changed to: info(at)ricardopilon(dot)com
All Rights Reserved ©2023 Ricardo V. Pilon
[i] Garner, C. (2022) ‘NDC Offers-Offer’.