A while ago I was flying across the Atlantic in a half full $250 million Boeing 747, wondering how it all worked financially. Multiplying the few hundred paid (round trip!) by passengers and 20 years didn't seem like it would pay for the plane let alone crews, fuel, maintenance and everything else. I even visited an airline once on business, and asked an employee during lunch break how an airline actually makes a profit. They did not know!
So here I am going to answer that, and also show the parallels to the software industry. The sources listed at the end include where I have picked up much of this information over the years.
Unless otherwise stated, numbers given are for 2019. They are general ballparks for mainstream passenger airlines, with some variance throughout the industry, and US centric. The numbers for specific airlines and aircraft of interest to you are usually publicly available.
There are two primary parts to the business:
- Operating an airline
- Making planes
The software business often has companies that both make software for distribution to all, and then separately operate that software as a service. The aviation business has been separated for almost a century.
Operating an airline
Simple: You spend vast quantities of people, money, and time. You will also outsource a lot. In return you will get back slightly more than you spent. In numbers it may cost you 12.4 cents per seat per mile flown, and you get back 12.6 cents per seat per mile, averaged across your entire operation.
The single most profitable thing is flying a full load of paying passengers, the bigger the plane the better. Until the 1990s a load factor of 65% was considered good. These days 90%+ is the target and that is usually the break-even point for a low cost carrier. Not filling your plane will lose you money, the bigger the plane the worse the loss.
The good news is that most expenses are proportional to flying time. For example the flight crew, cabin crew, fuel, maintenance, ATC fees etc are based on flight hours. Those expenses scale up with the size of the plane. Planes are flown for 8 to 12 hours a day, with bigger numbers being preferable.
Aside: Paying for planes
You won't be shelling out $250 million for a 747. The list prices were always aspirational, just like in enterprise software sales. The planes become a monthly payment, with lessors handling turning the big price into smaller monthly ones.
Based on this posting you can get a rough idea of what it cost for new aircraft in 2019. The prices go a lot lower for used/older. The number of seats varies by airline (eg business class seats take more space, there may be more or less galley space depending on flight lengths, there are denser slimline seats and less dense more comfortable thicker seats). Each aircraft also has sub-models offering incremental seating capacity at an incremental price.
With a software lens, it is also an enterprise sale. When someone spends tens of millions per plane, and usually buys many of them, there is a complex sales process. There are even legendary salesmen.
This is a giant optimization problem that plays out over months and years. You have to figure out what tradeoffs to make, and constantly update them while your competition do the same. Airlines have staff for which this is their job.
You could schedule flights and turnaround time for the duration they usually take. But any delay then affects operations later in the day since aircraft and crews aren't where they should be causing cascading problems. Making things looser by adding padding gives more buffer should problems happen, but then those same planes and crews aren't making you any money. Worst case you may end up doing 3 flights a day with planes when you could have done 4, and your competitors may be doing 4, making a third more revenue and providing a better schedule.
Tightening things up is a great way of making the business more efficient, until events exceed your spare capacity (time, crews, planes, parts etc). That usually results in cancellations, irate passengers, and negative media coverage. The spare capacity has a cost too, especially as it isn't used most of the time.
- Routes x Frequency
You want to serve as many places as possible to have a broad customer base. They won't want to split trips across multiple airlines. Travellers that are willing to pay more for tickets (eg business) also want more frequency so less of their time is spent waiting.
A common approach is to use smaller aircraft to feed passengers to larger hubs where they can be combined onto larger aircraft. But travellers willing to pay more want direct flights.
- Fleet complexity
You can get aircraft for virtually any number of seats (eg 20 seat increments from 70 all the way to 550). That means you could operate the perfectly sized aircraft on each flight. Crew are certified for certain aircraft models, maintenance varies, engines vary and overall you become less able to make changes.
Some airlines avoid the complexity by only operating one type of aircraft which makes it far easier to move crews, maintenance, spares etc around as needed. Others embrace the complexity by being able to put the perfect aircraft on each route.
- Fleet age
New aircraft are the most expensive to pay for and you'll have to work them hard to cover that. You do get to customize the cabin easily, making for a better onboard experience. Maintenance is also a lot less. (Replacing the worn out cabin in a 550 seat A380 costs about the same as a new 150 seat B737.)
Older aircraft are a lot cheaper, so it is easier to fly them only when it is worth it. But you'll have a more tired cabin. Maintenance costs also go up, and reliability will go down (a little). They will cost more to fly due to being less fuel efficient.
You'll notice some airlines that brag of youthful fleet get rid of planes at about 6 years old. That is when a heavy maintenance check (D Check) is done that involves taking almost the entire plane apart, checking everything, and putting it back together.
- Different offerings
You will not succeed if you charge every passenger the same amount. The standard is to charge more the closer to departure. It is common to have different seating classes, but you need to get the ratios useful for the routes aircraft operate - eg you want business and economy class to be full, not just one and flying empty seats for the other.
The easiest is charging for things that don't require changing the aircraft, like food, priority boarding, wifi, baggage etc.
You will never be able to handle everything yourself. For example if you operate one flight a day to an airport, then it won't make sense to have full time check in staff, full time maintenance, full time luggage handlers, full time cleaning staff etc.
Unless you have a lot of a certain aircraft, it won't make sense to do heavy maintenance yourself.
But outsourcing is more expensive - you are helping another company make money. The airlines outsource a lot of things to each other.
- Bonus: Freight
- A silver lining is carrying freight in the hold of passenger aircraft. About 90% of air freight used to be carried by passenger aircraft. They already fly where people go and have a timely schedule, so putting unused baggage space to work is pure gravy. It can also be what makes a flight that doesn't have a full passenger load still be profitable.
Simple: You spend vast quantities of people, money, and time. You will also outsource a lot. In return you will get back more than you spent, eventually, if the aircraft programme is successful.
Lines of code is a useful but very imperfect metric for software. (It does correlate with effort, complexity, bugs, functionality etc though). The equivalent for aircraft is weight, and that is how aircraft size is often measured. Weight has to be added to carry fuel (how far you fly), to contain passenger seats, and for aircraft elements like wings, landing gear, pressure vessel, catering etc. And more weight means more expensive to manufacture, design, purchase and operate.
The most important part is the manufacturing stage. The more you do something the better you get at it, improving efficiency. The standard way of measuring this is to compare the cost to produce unit number n with unit number 2n - for example 10 vs 20, 50 vs 100, 500 vs 1,000. Aircraft manufacturing is around 77%.
An example were estimates the first Boeing 787 cost $2 billion to make. The machines had to be made, machines to make those machines, staff trained, procedures worked out, mistakes detected and prevented in the future etc. It would take close to 1,000 planes manufactured at that 77% improvement before the manufacturing cost meets the sale price listed earlier!
It is a careful choice of how much of the design and manufacturing to outsource. It isn't feasible to do all of it yourself. Outsourcing to specialists reduces your effort, reduces your control, but they also expect to get more of the rewards. The Boeing 787 programme tried significantly increasing the amount of outsourcing, and is a good case study.
Engines are purchased separately from the aircraft. There is no standard fitting between the aircraft and the engine, and an airframe + engine combination is what is certified. You are stuck with the same engine model for the lifetime of a particular airframe.
The airlines prefer as much engine choice as possible, while engine manufacturers prefer as little competition as possible. They also have large investments to pay back. For example a deal with Boeing was made by GE to be an exclusive engine supplier for the Boeing 777-300ER model.
How does a new aircraft work?
To interest the airlines, you'll need to have a 15% fuel consumption improvement over what is currently available. Much of that improvment will come from improved engines, while the rest comes from improved materials (especially lighter ones) and aerodynamic tweaks (designing a new wing is very expensive and effective). It will however require actually designing and building the airframe and engines before the exact numbers are found.
Just like version 1.0 software will have "issues", the first aircraft off the production line will too, usually being overweight. (That reduces payload & range, and increases fuel consumption.) There is usually some sort of performance guarantee.
Those initial aircraft are going to have the most teething issues. And they are going to cost you the most to make, after having spent billions of dollars and many years. They will also have lower second hand values. That means launch customers will strike a hard bargain for the aircraft that cost you the most to make!
Bugs and small improvements are going to be found. Service bulletins (improvements) and airworthiness directives (affecting safety) are issued. There is careful tracking of each airframe since they could be implemented early during production, or later during maintenance.
A group of improvements to the airframe and engines can be bundled together into a "performance improvement package" - good for a percent or two in reduced fuel consumption. That makes for an easy upsell to customers whose aircraft haven't been manufactured yet. It is rarely sensible to retrofit existing airframes.
The aircraft manufacturer is now in a good position to make some good money. Every successful aircraft has been stretched - putting an additional fuselage frames ahead and behind the wing (to maintain center of gravity) and making space for a few more seats. The goal is keep everything else similar - avoiding new crew training, different maintenance etc. The airlines like it too - if you are flying a route with 200 seats that you routinely fill then the same plane slightly longer and slightly heavier with the same crew and 220 seats makes things easy. As an example wikipedia lists the 747 derivatives doing just that.
While software has it's versioning, aircraft have a different convention. Using the Boeing 747 as an example:
747: Refers to all aircraft of this family
747-100: The first model produced. The second was 747-200 etc. It isn't always the case that the model starts at -100 - eg if the second is expected to be a smaller aircraft then models may start at -200 with a -100 coming later.
747-436: While the model is conventionally referred to as the -400, they are different for each customer. The -436 is what British Airways had because of their specific engine and other choices like how the cabin is configured. There needs to be plumbing for toilets, electrical power for the galleys, and often choices about space being used for bags or additional fuel tanks. One documentary I saw years ago explained how customers could choose whether the clipboard clasp on the captains controls could be at the side or on top. When spending millions, the customer gets to decide!
Rewriting from scratch
By far the easiest thing to do is keep tweaking existing models. It costs about $2bn and 3 years to update and certify a new engine, and you may even be able to get the engine manufacturer to pay for that. The Boeing 737 has been going since 1968! That compares to the $20bn and 10 years for a clean sheet design, if everything goes well.
Eventually it gets too difficult - making the airframe longer becomes impractical, or needs longer landing gear which needs larger landing gear bays which forces all the other belly components to move. The efficiency improvements are harder to come by since you've done it several times. Rewriting from scratch will fix all these and more, but you won't know for 10 years. It is a difficult complex decision, just as with software.
- Leeham News and Analysis
- Excellent coverage of airlines and manufacturers, with good in depth analysis.
- Skyships Eng
- Wikipedia has good textual pages for aircraft. This Youtube channel has discussion and video of commercial aircraft history and operations.
- Cranky Flier
- Covers airline operations. During the pandemic Cranky has shown how the airlines kept updating their schedules and routes. There are also interviews with airline executives, and airport operators.
- The Aviation Herald
- Covers daily operational incidents world wide. You get an idea of how often there are bird strikes, engine issues, tail strikes etc happen (about 4 a day).