by David Kaplan MARCH 2, 2020
This article originally appeared on Kambr Media – check out this link and more for the latest news on the intersection of commercial aviation and tech.
“Should I have point-to-point? Should I have hub connectivity? Those are all of the possible ideas. And as you can imagine, that solution space is humongous,” says Amadeus’ Renzo Vaccari.
In the words of Renzo Vaccari, the head of the Amadeus Sky Suite group, the promise is that Sky Suite represents a clear break from the tools airlines and their tech partners have used to solve the inordinate complexity and intense competition associated with ensuring on-time performance.
Vaccari is certainly in a position to make such a bold claim. He joined Amadeus from his previous role as SVP for Airline Solutions at Optym, where he worked on advancing the technology for planning and scheduling for the last six years. Before that he worked on addressing those issues in a variety of management posts at Sabre Airline Solutions from 1996 to 2012.
Sky Suite is composed of five modules: SkyMAX (schedule optimization), SkySYM (schedule simulation), SkyCAST (schedule forecasting), SkyPLAN (route frequency optimization), SkyWORKS (schedule editing and management).
The importance of network planning and forecasting may take on added importance at this point.
Consider the sudden chaos and panic that have hit airlines in the past two weeks. From the plummeting of demand to the spate of route cancelations in the wake of the worldwide Covid-19 Coronavirus epidemic, the ability to manage scheduling is more than just about grabbing a potential edge among competing carriers; it could help determine which brands make it through the global health crisis with the least possible damage.
We spoke with Vaccari in early February when these industry’s current situation was looking much more positive.
Kambr Media: There is an increasing amount of competition among technology providers that promises to improve network planning and scheduling. How do you view the challenge within the airline tech landscape?
Renzo Vaccari: We see a huge opportunity for the supply side, the airline side, and from the technology side. Air traffic continues to rise. What does that mean? More planes; more people in the sky; more congestion; and more complicated problems. It’s simply become harder to schedule. At the same time, technology is getting better. Algorithms, data science, and modeling are getting better. Now, there are new techniques involving artificial intelligence and machine learning that can help solve these bigger problems than past systems.
Things that were impossible 10 years ago are now possible. That really is what drives us. There’s a saying, “Everything is impossible until somebody does it.” That’s how we start. These are very exciting times, in particular, for network planning and scheduling, because we believe that there is a better way of doing this. We believe there is a better way of solving and building schedules.
Our idea was to invent a totally different way of doing this.
How does Sky Suite advance airlines’ ability to better manage network planning and scheduling?
We are using new technology and new models, so that the technology is actually suggesting ideas – that’s totally different than in yesterday’s world. In yesterday’s world, the analyst comes with all of the ideas. Then you use technology as a simple calculator in order to score the idea. The score determines whether this a good idea or a bad idea.
For example, look at the ideas that determine fleet assignments. Which fleets am I going to use for certain flights? What ideas could apply to routings, or what is the best departure time for my flight?
Should I have point-to-point? Should I have hub connectivity? Those are all of the possible ideas. And as you can imagine, that solution space is humongous.
Look at it from a large scale. Take an airline like Southwest. It has 700-plus planes, depending on when their 737-MAXs come back into operation. Southwest has thousands of flights per day. The possible number of departure times, fleet assignments, and aircraft routings could amount to billions of combinations. You know that the value proposition yesterday was: “We have analysts, they will come up with a couple of possible scenarios out of gazillions, and then let’s hope that it’s a good one.”
The question now is, “How do we leverage new technology?” We actually have the machines doing what they can do very well, which is go through lots of possible permutations and, instead of picking the best one out of five options, let’s pick the best one out of hundreds of thousands of different scenarios. What we see is, not surprisingly, we see that that’s a better way of coming up with smarter schedules. That’s important, because the schedule defines the revenue potential of the airline.
Deciding where to fly, what to fly, when to fly, building that schedule, is going to define the potential for revenue management for pricing. Yes, you will be able to squeeze a little bit more revenue with the other commercial levers that you have. But your revenue power is certainly going to be driven by the schedule that you’re building to begin with.
So is it the more thoughtful use of machine learning that differentiates Sky Suite? Obviously, machine learning has become a buzzword in all industries the last few years –
I won’t use buzzwords. I think a combination of machine learning and, well, we hate to use the term “artificial intelligence.” We call it optimization intelligence. It’s the ability to use optimization to be able to have the technology suggest ideas taken to the extreme.
We are talking about a new concept called clean sheet scheduling. You start from scratch and, it’s almost going to sound too good to be true, but it’s true. With clean sheet scheduling, you are not limited by starting with a previous schedule. You are not limited by just making incremental changes. You can throw everything up in the sky and then rebuild your schedule and then have a very different schedule from month to month.
Sky Suite is made up of several impressive modules. Before we dive into each of them, where does clean sheet scheduling fit in?
It is part of SkyMAX, which is the schedule optimizer. It also runs in a traditional incremental mode, because there could be airlines that are just not quite ready to go ahead and make dramatic changes from schedule to schedule. Running it in incremental mode gives airlines the option to make more marginal changes for consistency between schedules, which is a more critical factor.
What we’re seeing with our customers is that if your demand changes significantly from month to month, from season to season, then there is significant incremental revenue to be able to optimize your schedule and have schedules that could be very different from month to month. That’s one of the key differentiating capabilities.
The other key differentiating capability is that you are not just trying to build schedules that are going to be good from a commercial perspective, but actually from a more holistic perspective. It allows you to look at schedules both from a commercial side but also from an operational side. And by operational, I specifically mean on time performance or reliability.
Those two perspectives compete with each other. Coming up with reliable schedules is not difficult. I can build a very reliable schedule because I add padding everywhere. I pad my block times. I pad my ground times. I just add a lot of slack in my network and then my own time performance is going to be very good.
But then the problem is that if you’re adding a lot of slack time in your system, that means the planes are on the ground for a long time. And that is not going to be the most profitable schedule. For us, you don’t have to choose between profitability and reliability. You now have the technology that can solve both at the same time. And so now, instead of having to add as slack blindly across the system, now you can be very laser-focused, and you can simulate what the on-time performance is going to be for these schedules.
We have been able to develop simulation capabilities that can predict OTP (on-time performance) within 1- to 2 percent.
What are the other modules that comprise Sky Suite?
The Sky Suite is a combination of five capabilities, five different core models. There is SkyPLAN, which is an optimizer for helping our lines decide where to fly. It identifies which markets to fly, and how many services per day, or per week, in each of those markets to maximize revenue. It helps draft your route strategy and it looks at your services. It looks at the competition, at market sizes, and at how many planes you have. It’s looking at all of those components and then it basically helps you decide where to allocate your planes or your block time to maximize revenue.
Then there’s SkyMAX, which is our schedule optimizer. This is the optimizer that now actually builds a schedule for you. SkyPLAN doesn’t build a schedule, it’s just giving you a frequency plan, a list of markets and how many services per market. You still need to translate that into an actual schedule with actual departure times, actual routings, actual fleet assignments. SkyMAX does that. Once you have the schedule, then you run it through SkyPLAN, which is our simulator. What it does is simulate the expected on-time performance. From there, it can suggest changes to optimize that on-time performance.
SkyCAST, which is our forecasting system, calculates what is going to be the expected traffic on all of your flight legs. The difference in SkyCAST with everything else in the marketplace is we have been able to invent some new models around passenger choice models and how passengers choose flights. We have been able to add some variables that didn’t exist before, such as price elasticity.
Nobody has really been able to figure that out in terms of how to match the right price to the right passenger. Everyone recognized that price is a factor, but no one’s figured out how to model that mathematically.
That’s where the price elasticity curve, and what data to use, and where new data that is available through Amadeus helps us do that. Now we can go away from using the traditional MIDT (Market Information Data Types) and we can get a lot more granular. We can look at shopping data, such as where a traveler is actually looking at shopping requests.
Can you describe that process and the underlying modeling approach for SkyCAST?
You have a passenger, the request, the point from origin O to destination D at this time. They are presented multiple options potentially with different types of services, different prices. From there, an airline can start building a profile of how passengers choose flights, including price.
Does Amadeus bring anything to the table in terms of data inputs to such a tool? Do you have plans to partner with any other third parties to enhance data input options for carriers in different regions of the globe?
This is just the beginning for us. Yes, part of the attractiveness of Amadeus, from the partnership with Optym and now even more so with the acquisition, is access to all of the data that they have. We think it’s just scratching the surface. There are, for example, some truly interesting insights around telecommunication data that is going to further help us understand market sizes and the flows of passengers between these markets.
From a forecasting perspective, the accuracy that we were able to get with the new algorithms, with the new data compared to what’s currently available in the marketplace right now is significant. Typically, you’re talking about error levels in the plus or minus 15 percent range as the best you can get.
SkyCAST is able to produce forecasts that are 7.5 percent. So SkyCAST is able to produce forecasts that are twice as good.
It’s not a marginal improvement. It’s not that you’re going from 15 percent to 13 percent. Even that probably would be significant. But you are able to make a major improvement in what’s traditionally been available. That translates directly into revenue. We’ve done empirical studies where we can show, both for network planning and revenue management, that a 10 percent improvement in accuracy translates directly to a 1 percent improvement in revenue. Those are some very interesting numbers.
Previously, you mentioned that this tool represents the ability to address a $400 million market potential. Can you unpack that 400 million number? Is that an internal estimate?
It’s slightly two different things. The $400 million is actually the airline spend in this space. Airlines on an annual basis spend about a $400 million in network planning and scheduling technology.
And yes, this was from an internal study of ours. We’ve done the internal study using two different criteria. They both converge around the same amount. So when you look at market share for the other vendors in this space, and then you do inversion and you look at what is their market share and what their revenue numbers are. Then, you can calculate roughly what is the total market size.
The other way of doing this is you can look at total number of passengers boards (PB) and then multiply it by an average price per PB in this area of network planning and scheduling. And then using both methodologies you come to roughly the same number.
In terms of the market segment most suited to using Sky Suite, this seems like a rather sophisticated, even complex set of solutions. Does an airline need an Operations Research group to get full benefits from the suite?
No, an airline doesn’t need an Operations Group to run Sky Suite. It is true that the underlying models in the Sky Suite are very advanced and very sophisticated. But then we have packaged that in a very simple, easy to use UI.
Let’s take the analogy of a car. A car has a very complicated engine. But I don’t have to be a super mechanic to be able to drive it, because I have a nice UI and interface and that allows me to work with the system.
I will say that it’s not just the technology and the tools either. It’s really the combination. The magic happens when you combine the tools with a person and the data. The analogy that I like to use is a Formula 1 car. It’s not just a car – you also need a good driver and you also need good fuel. In this case, the fuel is the data. The car represents the tools. And then the people, the analysts, are the drivers.
What about carrier scale? Does it make sense for smaller carriers, or is there a network size threshold where Sky Suite begins to make more sense from an ROI perspective?
We find that from a value proposition, this technology is really geared to your top 100 airlines. You need a certain scale to make the economics work. If you have hundreds of planes, then you certainly need some technology, because it’s just humanly impossible for people to process all that information manually. Even if you have a hundred people, a thousand people, it’s just impossible to go through all of the possible combinations in the relatively brief amount of time an airline has.
The flip side, if you have five planes, then I can do this problem in my head. I don’t need very sophisticated technology to, you know, I don’t need a bazooka to kill a fly.
Do you envision a “Sky Suite Lite” for some mid-tier carriers?
We do. Very, very much so. To further clarify my earlier point, it depends on the capabilities as well. We have capabilities like SkyMAX, SkySYM, SkyWORKS, and SkyCAST. Those are the more advanced decision support and optimization intelligence capabilities. We have the basic schedule editor, SkyWORKS, which is just a platform to go ahead and view, edit, manipulate, make changes manually.
The SkyWORKS system, which is more our core base schedule editor, could actually apply to everyone. All airlines need to be able to visualize their schedule. So we imagined having a “lite” version of SkyWORKS and then even a “lite” version of SkyCAST. We believe everybody needs to be able to do good network planning. But not everyone needs all of the full power.