How do airlines build their schedules?
Including a look at first quarter selling schedule changes made in January
We’ll tackle the subtitle first—it will serve as a useful backdrop for the titular question. We’d covered changes made in Jan 8 and Jan 15 updates in some depth, when airlines’ cuts spanned the entire first quarter (including extraordinarily close-in changes to January); by the Jan 22 schedule load, January and February schedules showed signs of having settled, though airlines continued to hack at March. While the last load of January (on the 29th) delivered more stability for the first two months of the quarter, airlines remained prejudiced towards March, removing nearly 7,000 flights. Rather than a focus on week-over-week changes, however, let’s consider the cumulative work done by Network Planning teams last month.
Across 4 schedule loads in January, airlines cut more than 60,000 first quarter flights. Impressively, there were 7 cases where an airline reduced a month’s selling schedule by at least 5%, including American’s nearly 9% March reduction. We were very intentional in our use of the word impressive: airlines’ schedule construction process resembles an assembly line and the closer a schedule is to fully-assembled, the more difficult it becomes to implement changes to its base architecture. While COVID has compressed schedule construction timelines, in most cases, March should have been fully-assembled—or very close to—by the start of January. Removing a half point of frequency could be accomplished in a few hours when the schedule is in its nascent stages (aided by automation); once a schedule is fleshed out, that same half point reduction is likely to take days (with less opportunity to employ automation).
How does a schedule flow through Network Planning?
Before we jump into the assembly line, we need to define who Network Planning encompasses. In a lot of cases, the organizational name—and the title for its senior leader—belies the fact that there’s two sides of the Network house. Planning conceives a schedule strategy, guided by economics and competitive influences; scheduling develops and implements planning’s vision, constrained by a number of variables (that we’ll cover later). Think of a puzzle: planning teams produce the picture featured on the front of the box and scheduling is charged with putting the pieces together. Invariably, compromise is required to finish the puzzle—planners will need to smudge or blot out parts of their picture while schedulers bend some pieces. This compromise produces some natural tension, as schedulers are predisposed to believe planners paint unrealistic pictures. For their part, planners have been known to accuse schedulers of being uninventive puzzle builders.
Because the various assembly line stations are most analogously staffed by schedulers, we’ll set aside planning considerations for a later explainer. Flight schedules are roughly defined by calendar month, spend approximately 30 days at each station and typically start in the hands of the Futures team. Futures normally works on a schedule that will fly in 4-6 months, i.e. we’d bet they’re currently working on the June flight schedule at most airlines. Futures doesn’t necessarily start from scratch: they may take a parent schedule off the shelf (i.e. generically built for a peak, shoulder, or lull period) or use the previous month’s schedule as the first version. Importantly, in this first stage of a schedule, analysts—and in many cases their leaders—are working in a single, base-day mode. This means Monday is not yet different from Tuesday, which is not yet different from Saturday. Futures may also aggregate1 some aircraft types at this stage, in which case a separate exercise may be required to fully “fleet” the airline.
It’s not until Futures hands off a schedule to the Day of Week team that unique plans are implemented for different weekdays (and holidays). While Mondays, Thursdays and Fridays will often bear a strong resemblance to that schedule’s base day, Tuesdays, Wednesdays and weekends can be substantially re-written. In terms of making unscripted schedule changes, a schedule going through the day of week process is an important bridge to cross. Why? Schedulers may be required to develop a unique solution for each day of the period—there’s a multiplicative effect with respect to the amount of time required to implement changes. A schedule matriculates from Day of Week to Current, where it’s combed over for the sake of operability; pilots and flight attendants will also be “bid” or assigned while the schedule is in Current’s shop. Somewhere around 7-10 days from departure, the schedule will be handed off for the last time, from Current to the operations team.
Even pre-COVID, Network teams were required to navigate a fluid environment and it was not unusual for Current to implement root changes. Perhaps delays to a heavy maintenance check or an undersubscribed pilot training class invalidated assumptions that Futures had made 2 months ago. However, the scale of these first quarter changes made during the last month is difficult to convey. That January selling schedules were adjusted in January is especially remarkable, as crew bids having been completed is perhaps the most restrictive scheduling constraint.
What constraints do schedulers consider?
So we’ve referenced scheduling constraints a couple times now—what do we mean? These are the rules that schedulers must abide by when putting the puzzle together. Each of the below constraints will be considered from Futures to Current, however the further down the assembly line a schedule gets, constraints are liable to be applied with more granularity.
The most fundamental constraint is aircraft time. If a hypothetical airline wishes to operate 3 cross-country round trips per day, they’ll physically require at least two aircraft: each round trip entails approximately 10 hours of flying time (plus an hour of ground time, conservatively) and like people, planes only have 24 hours in a day.
Even if the same hypothetical airline dropped their plans for the third round trip, they’ll still need a second aircraft. While 20 hours of flying time physically “fits” in an aircraft day, it doesn’t provide for sufficient maintenance touch time. Regular, “line” maintenance requires 7-12 hours of ground time at least once per week. In a similar vain, airlines will often elect to set spare aircraft aside.
So you’ve fit your flight schedule to available aircraft time and ensured sufficient maintenance opportunities. Great—do you have enough pilots to fly the schedule? Pilot availability is governed by FAA regulations at a minimum (and often further constrained by collective bargaining agreements). Additionally, to ensure pilots are deployed efficiently, scheduling has to be attentive to the length of overnight layovers.
While the constraints we’ve consider thus far are of the regulatory or physical variety, Network Planning will often self-impose bank start and end times. To optimize connectivity, airlines will often fit arrivals and departures into narrow windows. Slots function in a similar manner—fitting operations into a defined window—though de-peak, rather than concentrate, activity. A handful of airports also have curfews prohibiting late-night and early-morning operations.
Alright! You’ve checked off aircraft time, maintenance touch time, pilot availability and bank compliance. Do you have enough gates to accommodate the schedule? Banking exacerbates the demand for gates, however gating problems are not confined to hubs. In a broader sense, schedulers also have to ensure that the total on-ground inventory of aircraft can be accommodated between gates and remote, hardstand parking.
Finally, airports are only equipped to handle certain types of aircraft (or certain regional carriers). In addition to airport equipage and training, airlines will self-impose aircraft type restrictions on flights between certain cities for marketing reasons (e.g. disallowing an aircraft that doesn’t have a first class cabin).
To our Network readers—what’d we miss? To our non-Network readers—is there a topic, team or constraint you want to read more about? We had mentioned a planning explainer, for which we would welcome a guest author. Pricing & Revenue Management also operates in parallel to Network Planning & Scheduling, though we’d definitely need a guest author for that explainer. Email us at tim@aerology.ai in any case!
For example, if an airline is in the middle of a project to change the seat configuration on the their 737-800 fleet, Futures may not breakout the new configuration from the old and instead schedule the entire fleet as the predominant configuration.