Somewhere in the twenty minutes before your flight pushed back, one of the pilots typed a number into a keypad on the centre console. It was probably two or three digits. It took about four seconds. That number decided how fast you are about to fly, how much fuel the airline will burn getting you there, and whether the aircraft will bother trying to claw back the eleven minutes it lost at the gate.

It is called the cost index, and it is the most consequential number in commercial aviation that almost nobody outside the industry has heard of. It is also, as Boeing gently noted in its own technical magazine, a tool that a lot of airlines do not really use properly.

A ratio, not a speed

Start with what the number actually is, because the name is unhelpful. The cost index is not a cost and it is not an index of anything in the ordinary sense. It is a ratio of two costs. Boeing writes it as the time-related cost of an airplane operation divided by the cost of fuel. Airbus writes the same thing as C_T over C_F.

What makes it useful is that the flight management computer does not need to know either cost in absolute terms. It only needs to know their relationship. Tell the aircraft that an extra minute in the air is worth thirty kilograms of fuel and it can work out, for this weight, this altitude and this wind, precisely which Mach number minimises the total. It will then fly that speed in climb, in cruise and in descent, adjusting continuously.

The units are where the first real confusion starts, and it is a genuine trap rather than a pedantic one. Airbus expresses cost index in kilograms per minute. Boeing’s equation is dollars per hour divided by cents per pound, which resolves to hundreds of pounds per hour. Since one kilogram per minute is 132.28 pounds per hour, Boeing numbers are about 1.32 times Airbus numbers for the same underlying economics. Airbus prints both columns in its own documentation and the factor holds to the rounding.

So a cost index of 50 on an A330 and a cost index of 50 on a 777 are not the same instruction. They are not even close, because the scales differ as well as the units. This matters enormously to an airline running a mixed fleet and not at all to the aircraft, which only ever sees its own number.

The thing almost everyone gets wrong

Here is a piece of aviation folklore you will find repeated across forums, training notes and a good many explainer videos: the A320 uses a cost index range of 0 to 999, and the A330 uses 0 to 99, because they are different aircraft.

That is not what Airbus says. In Getting to Grips with the Cost Index, the manufacturer attributes the split to the flight management system vendor. Sperry and Honeywell systems scale 0 to 999. Smiths systems scale 0 to 99. The brochure captions a figure with the phrase "scaled 0 to 99 or 999 (depending on FMS vendor)" and leaves no room for ambiguity.

The type-based version of the rule works most of the time, because a given Airbus model tended to be delivered with a given box. But it is a correlation that people have mistaken for a cause, and it falls over the moment you meet an aircraft fitted with the other vendor’s equipment. The scale belongs to the computer, not the airframe.

Boeing avoids the problem by publishing a straightforward table, and the most interesting entry in it is the oldest. The 737 Classic — the -300, -400 and -500 — runs 0 to 200. The 737NG family runs 0 to 500. The 747-400 and the 777 run 0 to 9999. The 757 and 767 are listed as 0 to 999 oder 0 to 9999, because it depends on the flight management computer build standard. Even Boeing’s own table needs an "or" in it.

Cost index zero is a floor, not a minimum

This is the second thing people get wrong, and it is the more interesting error because it reveals something real about how aircraft work.

If you enter zero, you are telling the aircraft that time costs nothing. Boeing states that for all models this results in maximum range airspeed and minimum trip fuel, a schedule that ignores the cost of time entirely. Airbus says the same thing in its own vocabulary: with zero cost of time the flight management system flies at Mach for maximum range.

The intuitive reading is that CI 0 must be the slowest the aircraft will go. It is not. Maximum range cruise is the peak of the curve of distance travelled per kilogram of fuel. Slower speeds certainly exist — holding speed, green dot, maximum endurance — but every one of them burns more fuel per mile than MRC does. Going slower than cost index zero costs you fuel on the trip, not saves it. The Associated Press put this well for a general audience during the 2008 fuel crisis: below a certain speed, which varies by aircraft type, an aircraft’s fuel usage can actually rise.

At the other end, maximum cost index means fuel is free and only the clock matters. Boeing goes to VMO/MMO in climb, cruise and descent. Airbus is more precise and slightly more conservative: whatever the model and the conditions, the ECON Mach is limited to MMO minus 0.02. For the A320 family that is Mach 0.80; for the A330 and A340, Mach 0.84.

The most-watched explainer on the subject, from an airline pilot, at a little over seven minutes.

Why the whole thing works: the curve is flat on top

Specific air range is the distance an aircraft covers per unit of fuel — true airspeed divided by fuel flow, expressed in nautical miles per kilogram. Plot it against Mach number and you get a smooth hump with a single maximum.

The useful property of any smooth maximum is that the slope at the top is zero. Move away from the peak and your range efficiency falls off only in proportion to the square of how far you moved, while your speed rises in direct proportion. Near the top of that hump you buy a great deal of speed for a very small amount of fuel efficiency.

The aviation industry has a name for exploiting this, and it predates cost index by decades. Long range cruise is defined as the speed at which specific range equals 99% of the maximum. Airbus explains the logic in one sentence that carries the whole idea: the 1% loss compared to maximum specific range is largely compensated by the cruise speed increase, due to the flatness of the curve.

How much speed does that 1% actually buy? Here it is worth being careful, because a figure of "3 to 5% faster" circulates widely and cannot be traced to any manufacturer document. Airbus publishes only the 1% specific range loss and describes the speed gain as "significant" without quantifying it. The best sourced answer comes from academic work: across six aircraft types, moving from maximum range cruise to long range cruise costs between 0.6% and 1.2% in block fuel and saves between 0.7% and 2.3% in block time. Less dramatic than the folklore, and the cruise segment gain is larger than the block figure because climb and descent dilute it.

Cost index is the general solution to the same trade that long range cruise solves with a fixed rule of thumb. And in practice the optimum Mach usually sits between maximum range cruise and long range cruise, which is to say the aircraft spends most of its life in the flattest part of the hump.

There is a corollary that airlines quietly rely on. Because the cost curve is flat near its optimum, getting the cost index slightly wrong barely costs anything. One study of the problem notes that variations in cost index near the optimum value have a negligible effect on total costs. This is why a carrier can run a company-wide default of "35" across an entire fleet, never revisit it, and not be visibly punished for it.

Nobody knows what an hour costs

Which brings us to the genuinely hard part. Fuel cost is easy: an airline knows to the cent what it paid per kilogram. The numerator is the problem.

Airbus itemises time-related costs as hourly maintenance cost excluding cycle-driven items, flight and cabin crew cost per flight hour, and marginal depreciation or leasing cost for extra flying — then adds that overtime, passenger dissatisfaction, hubbing and missed connections may apply and are airline-specific. Boeing explains why this list is so difficult to turn into a number.

Crew wages may be hourly or fixed. Engines, APUs and airframes may be leased by the hour or owned outright. Maintenance may be accounted by the hour, by the calendar or by cycles. Each item may therefore have a direct hourly cost or a fixed cost with limited or no correlation to flying time at all. Two airlines flying identical aircraft on identical routes can legitimately arrive at wildly different numerators.

Airbus is blunter than you might expect about why this does not get fixed. Much progress could be obtained, the brochure says, by having airline accountants look into the other time-related costs — but in practice it has been hard for flight operations departments to persuade their financial analysts to assess marginal operating costs, probably because the importance of the cost index is largely an unknown concept to their decision-makers. That is a manufacturer telling its customers, in print, that the wrong department is ignoring them.

Modern European data shows how far apart the plausible answers sit. Work done for EUROCONTROL puts marginal crew cost per minute at 7.8 euros for an A320 and 16.7 for a 747-400, with airborne tactical maintenance at 3.4 and 7.6 respectively. But the low scenario in the same tables is 0.00 for every single type. At the margin, some airlines’ crew cost genuinely is zero. That single fact explains most of the variation in cost index settings across the industry.

The numerator is mostly your passengers

The reason airlines care about the time term at all becomes obvious once you look at what a delay costs, and the shape of that cost is the most surprising number in this article.

For an A320 on a European en-route sector, the same EUROCONTROL reference values give a five-minute delay a tactical cost of about 260 euros. Fifteen minutes costs 960. Thirty minutes costs 2,420. Sixty minutes costs 6,650. A 737-800 and a 747-400 follow the same curve at different heights, with the 747 reaching nearly 20,000 euros at the hour mark.

Notice that a sixty-minute delay costs an A320 operator roughly twenty-five times a five-minute one, not twelve times. Delay cost is violently non-linear, because at some point passengers start missing connections, crews start running out of duty hours and aircraft start missing their next rotations. By the two-hour mark, passenger costs account for 80 to 90% of the total across all aircraft types.

That is the economic punchline of the whole subject. When a flight management computer is weighing an extra minute in the air against an extra few kilograms of fuel, the minute is expensive mostly because of the people sitting behind the flight deck door and what happens to their onward journeys. The cost index is, in the end, a number about passengers wearing a disguise made of kerosene.

The researchers who compiled those figures make the link explicit, noting that an aircraft may depart late and then attempt to recover all or part of the delay by flying faster than originally planned, using a higher cost index — with implications for maintenance costs under some contracts and, more importantly, for the predictability of aircraft trajectories from an air traffic controller’s point of view. Speeding up to make up time is not free and it is not invisible.

Why a headwind makes the aircraft speed up

Ask most passengers what an aircraft should do in a strong headwind and they will guess it should slow down and save fuel. The aircraft does the opposite, and the reason is elegant.

Airbus states the rule in both of its relevant brochures, in almost identical language. Tailwinds increase the specific range and lower the speeds; headwinds decrease the specific range and raise the speeds. Or, in the cost index brochure’s phrasing: headwinds command higher ECON speeds because of less exposure time to the wind, and tailwinds command lower ECON speeds, to let the winds work.

The mechanism falls straight out of the definition. The cost of covering a nautical mile has a time term of cost index divided by ground speed. In a headwind the ground speed is low, so that term is large — every extra minute aloft is buying you less distance, and the marginal value of going faster rises until the steepening fuel curve balances it. In a tailwind ground speed is already high, the time penalty per mile is small, and the optimiser slides back toward maximum range cruise and lets the atmosphere do the work.

Airbus is careful to note that this is a result of the cost index definition itself and not of any particular flight management system implementation. It also attaches an operational warning worth repeating: in managed cruise, pilots should pay particular attention to the Mach number in strong headwinds, especially with high inserted cost indices, since this could lead to significant cruise speeds.

What happens when fuel gets expensive

If cost index is a ratio, then doubling the fuel price halves it, and the whole fleet slows down. This is not theoretical. It happened, visibly, in 2008.

With New York Harbor jet fuel around $3.54 a gallon and up 73% in a year, Southwest added one to three minutes per flight and reported saving $42 million annually. JetBlue added just under two minutes on average for $13.6 million. United estimated $20 million through flight-planning software. Northwest added four minutes on its Hawaii routes for $600,000, and on a single Paris to Minneapolis service added eight minutes, bringing the flight to 8 hours 58, saving 162 gallons worth $535 and dropping the average speed from 542 to 532 mph.

He was right, and that is the point. Two minutes per flight is invisible to a passenger and worth $13.6 million a year to an airline. The economics of cost index are made entirely of differences too small to notice, multiplied by a very large number of flights.

The relationship has since been measured properly. Work published in Economics of Transportation in 2024 found that a 10% increase in fuel cost reduces flight speed by 1.2%, which at an average speed of 455 mph amounts to 5.5 mph. Slow, steady, and entirely driven by the price of kerosene.

Lufthansa’s own explanation of the cost index, in German, from one of its pilots. The only carrier-official video on the subject we could find.

So how much time does it actually buy?

Less than you would hope, and the returns collapse quickly.

Airbus publishes worked figures for the A319, A320 and A321 on a 1,000 nautical mile sector including takeoff, a step climb between FL350 and FL390, cruise and descent. Measured against a cost index of 0: CI 20 saves 3 minutes for 50 kg of extra fuel. CI 40 saves 5 minutes for 150 kg. CI 60 saves 7 minutes for 250 kg. CI 80 saves 8 minutes for 350 kg. CI 100 saves 9 minutes for 400 kg.

Read that sequence again, because it is the whole economics of the subject in six numbers. The first twenty points of cost index buy a third of the available time saving for an eighth of the total fuel penalty. Everything after that is increasingly bad value. Going from CI 80 to CI 100 costs 50 kg of fuel to save a single minute.

This is why a delayed aircraft does not simply "speed up and make it back". On a short sector there are perhaps nine minutes in the entire range of the setting, and the last of them are ruinously expensive. On longer sectors there is more to play with, but the shape of the curve is the same.

One more number, for the pilots reading. Airbus tabulates what happens if you simply fly 0.01 Mach faster than ECON in managed mode: on a 1,000 nm A320 sector that is 40 kg of fuel for 2 minutes; on a 6,000 nm A340-313E it is 900 kg for 10 minutes. One hundredth of a Mach number, and most of a tonne of fuel.

If you want the full twenty-five minutes, a line pilot walks through cost index on the 737, including what the number does on a real flight plan.

The four-second decision

There is a reasonable argument that cost index is the purest expression of what commercial aviation actually is. Every visible thing about a flight — the schedule, the fare, the connection you are running for — sits downstream of a trade between two costs, one of which is precisely known and one of which most airlines have never properly calculated.

The physics is generous. The specific range curve is flat on top, so the aircraft will let you buy a lot of speed cheaply and will forgive you for getting the number roughly rather than exactly right. The economics is unforgiving in a different way: the minutes you most want, at the top of the range, are the ones you can least afford.

And somewhere over the Atlantic tonight, an aircraft that met a 60-knot headwind will be flying a fraction faster than the one behind it, burning more fuel to spend less time in air that is moving the wrong way. Nobody on board will notice. The number in the box worked it out before they finished boarding.

Sources: Airbus, Getting to Grips with the Cost Index, Issue II, May 1998; Airbus, Getting to Grips with Fuel Economy, Issue 4, October 2004; Airbus, Getting to Grips with Aircraft Performance, January 2002; Bill Roberson, "Fuel Conservation Strategies: Cost Index Explained", Boeing AERO Q2 2007 (SKYbrary mirror); Cook, A. and Tanner, G., European Airline Delay Cost Reference Values, Version 4.1, University of Westminster for EUROCONTROL, 2015; Cook, Tanner, Williams and Meise, "Dynamic cost indexing: managing airline delay costs", Journal of Air Transport Management 15(1), 2009; Edwards, H.A., Optimisation of the Aircraft Cost Index for Air Travel Emissions Reduction, PhD thesis, University of Leeds, 2015; Brueckner, Kahn and Nickelsburg, Economics of Transportation 38, 2024; Associated Press, "Airlines slow down flights to save on fuel", 2008. Airbus cost figures are from a 1998 document: the methodology is current, the dollar values are historical.