Crew scheduling
No spreadsheet can hold every flight, crew licence and duty-time rule at once. We plan duties and crews together, by the rules, at the lowest cost.
- Free 30-minute call
- No commitment
- You own the code
In short: A crew scheduling optimisation builds duties for pilots, cabin crew or train drivers and assigns named people to them. Every flight or train is covered, every duty-time limit, rest period and licence rule is kept, and crew cost stays as low as possible.
The problem
On Thursday afternoon, the inbound flight from Lisbon is two hours late. With their evening flights, its crew would now be on duty longer than allowed. You need a rested replacement at the right base, qualified on the right aircraft. Every swap you try makes another duty later in the week illegal.
Crew scheduling for an airline or rail operator puts qualified people in the right seats at the right times. It has to cover every flight or service and keep a long list of duty-time rules. It’s widely seen as one of the hardest problems in operations research, the maths of planning.
Why crew plans fail when you build them by hand
No spreadsheet models crew scheduling, and nobody can hold it in their head. The rule set is enormous, and every duty you assign limits what is still allowed later in the month. One illegal duty can upset a crew member’s whole month of work, called a bidline. Mistakes show up fast and in public, as delays, cancellations and crews who run out of hours.
Why generic crew software falls short
Generic crew software is rule-based. It builds pairings, chains of flights one crew works before returning home. Each pairing passes a checklist, and then the software hands them over and stops. It has no objective function, no score for comparing two valid plans on cost. So it can’t tell whether its pairings are the cheapest valid set or just the first that passed. When no plan meets every rule, the software stops with an error message. Or it silently loosens a duty-time limit you have to keep.
What changes
Today, a late inbound flight means an evening on the phone. You swap duties by hand and check each rest rule against the rulebook. Our model recalculates the current plan around the delay. It proposes repaired duties and reassigned crews that keep every duty-time limit. You check the proposal and confirm it.
Crew scheduling optimisation: what you get
You check every plan against numbers you already watch: pairings needed, standby and deadhead cost, fairness and planning time. Standby is crew paid to wait for a call, and deadheading is crew flying as passengers to their next duty. Every roster keeps every duty-time limit before you publish it.
We test the model with your schedules and on your worst disruption days before anything goes live. Duties and monthly rosters come out of one model. So every monthly roster fits the duties it is built from.
- A model written around your own crew scheduling rules.
- Results measured in the numbers you already track.
- It slots into how you already work, and you can check every answer it gives.
How it works
Every shift, job and resource goes into one model, and your rules go in as hard limits. You get the best schedule that keeps all of them. What “best” means is your call: lowest cost, the fairest split or the fewest late orders.
Crew pairing: building the duties
The problem has two stages: first the duties, then the people. Crew pairing builds the duties. Starting from the published flight schedule, it assembles pairings. A pairing is a valid sequence of flights and duties one crew can work in a single tour. In between, the crew rests at an outstation, an airport away from base, or returns home.
Pairing is a covering problem: the pairings must cover every flight leg exactly once, at minimum cost. Each one must fit the maximum flight duty period, the longest a crew may be on duty, plus the rest and flying-hour limits. An airline crew pairing algorithm searches for the cheapest set of pairings: fewer pairings, with less paid waiting between flights. Those mean less standby, fewer hotel nights and less deadheading.
Crew rostering: assigning named crew
Crew rostering assigns pairings to named people, usually for a month. Base rules keep crew starting and ending close to home. Qualification tracking records who is licensed and current for each aircraft or route. Recency rules check who has flown enough lately to stay qualified. Fairness and preference rules settle who gets the unpopular roster and who gets the easy one.
Most operations solve the two stages separately. Then a pairing may need an aircraft type that too few crews at its base are qualified for. It stays open, and you look for a replacement at short notice. Our model plans both stages together, so every pairing fits the crews you have.
Duty-time limits in the model
Your duty-time rules usually apply over several periods at once. They cap duty hours per week, per two weeks and per month, each counted back from today. Flight hours have their own caps per month and per year. A rest often has to last at least as long as the duty before it. And crew need a minimum stretch of time off every week.
Staff agreements add more rules, and they differ between airlines, aircraft types and rail operators. All of it goes into the model as constraints it proves it has met. You stop adding up hours across a calendar by hand, and every proposal arrives already checked.
Worked example: the Lisbon delay
Back to Thursday. The crew’s duty was planned from 08:00 to 18:00, ten hours. Say rest must last at least as long as the duty before it. Next morning they report at home base at 06:30, after 12.5 hours of rest, which fits. Then the inbound runs two hours late. Duty now ends at 20:00 after 12 hours, so rest must last 12 hours too.
The earliest valid report is 08:00, and the 06:30 departure has lost its crew. The replacement has to be at that base, rested, qualified on the type and inside their own 7-day and 28-day totals. The model checks all those knock-on effects at once, proposes the cheapest valid repair, and you decide.
Disruption recovery on rail
A train driver scheduling algorithm faces the same task. One late train can ripple through a whole day of driver duties. The working-time rules differ from aviation, but the steps are the same. First you build the duties. Then you assign them to named people. And when a train runs late, you repair the plan.
How we build your crew model
Mixed-integer linear programming (MILP) writes the plan as yes-or-no choices and searches for the cheapest one. Constraint programming rules out impossible combinations rule by rule. The model proves that it keeps every duty-time and rest limit. We plan pairing and rostering together. The plan keeps every rule, covers every flight and stays fair, at the lowest cost.
Common questions.
Something else? Ask us directly.
Yes. Many operators fly under more than one rule set, or with different agreements per base or fleet. The rule sets are similar in parts and differ in the details: how rest is counted, which periods the maximum hours apply to and what counts as duty. A model for all of them needs each rule set written separately. Each crew member is then checked against the set that applies to them.
Yes. In the US, Canada and Australia, seniority usually decides who gets which monthly roster. Seniority is how long someone has worked for the airline. Crew members hand in their wishes for the month. The industry calls this bidding. Many European airlines roster without seniority and aim for equal workloads instead. A model can follow either method. With seniority, it meets the wishes in order: whoever has been there longest chooses first. Without seniority, it spreads the work as evenly as it can across everyone.
Yes. The structure is the same: build duties that cover every train, then assign them to named drivers, then repair when something goes wrong. What changes is the rule set. Train drivers have their own working-time rules and agreements. They also need to know the route and the type of locomotive they drive. We take those rules from your documents, never from a template.
Sometimes it can, and if that’s the case, we’ll tell you. Standard software is built for an average operation, though. Yours runs on its own rules: your contracts, your exceptions, the things you know by heart. A custom model writes those rules in, and that’s where the better plan comes from.
You don’t need either to start. We build the model from the data you’ve already got, even if it lives in three spreadsheets and one person’s head. You get a system you run yourself. We train your team until they no longer need us.
You see a first result with the prototype, before you commit to anything bigger. We start with a free 30-minute call about your operation and the model that would fit it. Then we build a working prototype and run it on your numbers.
Similar pages.
Terms on this page
Every abbreviation explained in plain words, in our article “Metrics and methods”.
Tell us what you’re planning.
In a free 30-minute call with a specialist, you find out whether optimisation can help.
- Free, no commitment
- You talk to a specialist directly
- A clear go or no-go