Many years ago, I used to work as a TPF mainframe software developer, building applications for one of the leading global PSS providers at that time. Over the years I have had the privilege of working on some ground-breaking projects. When I first started in the mid-nineties, we were putting in place API layers for web services to power some of the first airline e-commerce platforms. In the early noughties, I was involved in the integration of one of the first origin and destination (O&D)-based revenue management systems, promising to deliver incremental revenue gains of 1-2% for airlines. This was, and still is, big money for any carrier.
Around 15 years after this project, in my role as solution architect I was responsible for integrating another airline with this same RM application. Not surprisingly, considering the pace at which the airline industry evolves, this integration was more or less identical to the initial implementation, although with a different PSS provider. Every night, a dump of booking, inventory and schedule data is pushed to the RM application which ingests this data along with numerous other files containing flown ticket data and who-knows-what else and begins running its nightly optimisation processes. Around eight hours later, new steering controls, bid prices and so on are pushed back into the reservation system and the process is complete for another day. Outside of this, ad-hoc changes may be triggered for a flight, either manually or automatically based on certain events. Essentially though, for almost all of an airline’s network, each flight goes through this process once a day.
Optimising the price of every seat on every O&D of an airline’s network is a very complex process, and back in the eighties when the first airline RM systems were implemented, this daily cycle was all that was technically possible. The enormous computing power needed was both expensive and scarce, and only available to airlines with deep pockets (we carry more computing power these days in our pockets!). Pretty much every airline RM system still works this way today: batch data is downloaded from booking systems (i.e., the PSS), optimisation processes run, and the output is uploaded into the airline’s pricing and inventory control systems (usually PSS). However, the (technology) world has moved on since then: computing power has become much more affordable, and the growth of cloud technology has made this available on demand and instantly scalable. At the same time, the volume of airline shopping transactions has increased exponentially in the last decade or two. Airline products have also become diversified and more complicated, with the advent of de-bundling components of the air ticket (seats, bags etc.). Markets have become more competitive, with demand exceeding supply in most cases. Considering all these factors, one must consider whether the RM approaches still used today are effective.
In one regard, the answer to this question is clearly yes: the RM methodologies themselves. Many clever people have dedicated their lives to perfecting the algorithms used to forecast demand based on all manner of data sources, statistical methods, and highly complex algorithms. These continue to adapt to the new ways in which airlines price and sell their products, although this is still predominantly limited to the air fare only. However, it could be argued that the manner in which these powerful algorithms and calculations are applied is somewhat outdated, considering the technological capabilities available today. Let’s consider an airline carrying 50 million passengers a year in a typical hub-and-spoke network. Using some schoolboy mathematics, this might give an average of around one booking created every second, give or take a few. For reference, Amazon gets something like 18 orders per second[1]. Assuming the airline is using O&D-based revenue management, this potentially means that the demand on a significant portion of the network has changed – and therefore of course the price. But these incredibly dynamic changes are not ingested by the forecast algorithms until the RM machines get their batch files to churn through and deliver new demand forecasts hours later. Of course, airline pricing is much more complicated than most products sold through online retailing, where prices are (relatively) static, but does that not mean that it is even more important that airlines stay on top of pricing and adapt in real time?
What is holding us back?
So why don’t airlines do this non-stop, 24x7x365? Well, the answer is the same as for many questions in the airline world: silos. Way back before many of you were born, there was just PSS – schedules, inventory, PNRs and tickets (eek!). The RM systems were bolted on using big interface files. But today’s computing world looks different – we have real-time integrations, artificial intelligence and machine learning engines that never sleep and enough computing power to run the numbers over and over again and get the results instantly. With the advent of offer and order management systems, we are also a goldmine of offer, pricing and conversion data that is just waiting to be tapped into. Sending a dump of booking data can tell you a lot about what was sold, but nothing about what was offered or who asked. Unlocking the value in this data and understanding what it tells you is the key airline retailing – offering the right products in the right channel at the right price.
Traditionally this data has been difficult to interpret – EDIFACT messages, tickets, fare base codes, RFISCs, RBDs all in cryptic formats. NDC and ONE Order bring some standardisation to these key sources of data, but we need to work harder to break down the silos and truly start working with offers and orders (instead of just bolting them onto our legacy systems).
Indeed, this issue is not only to be found within the RM domain. Many of the initiatives in the industry at present are reliant on removing these silos in the end-to-end chain of distribution. Instead of a set of standard integration points based on interfaces from the 1990s, a dynamic and real-time exchange of key data is needed to be able to make offers that are truly relevant, targeted, and likely to lead to conversion. The flow does not simply end with the completion of a booking. Real-time delivery of sales into financial accounting can simplify settlement and revenue recognition. Real-time operational data can drive automated, proactive service recovery in case of disruptions – a task today that often requires extensive manual intervention. For far too long, as an industry, we have looked at these barriers individually – and indeed, in the execution this is the way forward. However, we must also to step back and look at the flow in its entirety – offers, orders, service delivery, payments, financial accounting, RM, customer management and so on. This transformational journey will involve many steps along the way, but without seeing the big picture, the course cannot be plotted.
At Travel in Motion, we are passionate about driving this change forward – let us share our expertise with you and help guide you on your transformation to a world of offers, orders and airline retailing and unlocking the value in that vast amount of data.
[1] https://landingcube.com/amazon-statistics/
This post has been published in collaboration with Terrapinn.
(Nick Stott, 5. April 2023)