Scheduled Seats Against Flown Seats: Why Published Capacity Reads High

Every capacity figure quoted in a route study, a press release or a board pack begins life as a schedule file — a statement of what an airline intends to fly. Nothing in that file records what left the gate. The gap between the two is not random noise: it is structural, it has a direction, and it is nearly always the same direction. Here is where the seats go, and how to take them out before the number turns into a decision.

A schedule is an intention, not a record

Commercial schedule data has a common ancestry. Airlines publish their intended programmes in the industry’s standard schedules format — the structure defined by SSIM — and aggregators collect, normalise and resell it. Whatever the interface, the underlying row is the same: an operating carrier, a flight number, an origin and destination, a period of validity, a pattern of operating days, a departure and arrival time, an aircraft type code, and a seat count.

Everything a capacity report contains is arithmetic performed on that row. Multiply the seat count by the number of days the pattern fires inside the period, sum by market, and you have « seats ». Multiply again by great-circle distance and you have available seat kilometres. The arithmetic is unimpeachable. The input is a plan.

There is no field in that record for whether the flight operated. There cannot be: the record is written before the flight exists. A schedule file is the aviation equivalent of a signed order book — useful, forward-looking, and systematically different from revenue.

The subtraction chain between filing and flying

Between the filed row and the seats that physically moved, several independent mechanisms remove capacity. They do not cancel one another out, because almost all of them subtract.

Passengers boarding a small white airplane on an airport tarmac at dusk.
Scheduled capacity often exceeds actual flown seats.

Cancellations. Commercial cancellations in thin periods, technical unserviceability, weather, air traffic flow restrictions, crew shortfalls, industrial action. A cancellation removes the whole row for that day: every seat, every kilometre, and the departure itself.

Equipment substitution. The flight goes, but on a different aircraft than filed. Downgauging is the more common direction — a smaller sub-fleet variant covering for an aircraft out of service, a widebody swapped for a narrowbody in a soft month. The departure survives; part of the seat count does not.

Configuration mismatch. Schedule data carries an aircraft type code, and many analyses convert that code to seats with a default table. Real cabins vary by operator, by sub-fleet, and inside a sub-fleet by retrofit status. A high-density leisure configuration and a long-haul-configured aircraft of the same type can differ by a third or more. Nothing in the type code signals which one flew.

Marketing duplication. This is the single most common error in a naive count. A schedule feed contains both the operating carrier’s row and the marketing rows of every partner selling the flight under its own code. Summing carriers without collapsing to the operating flight counts one aeroplane several times. On heavily code-shared long-haul markets, the inflation can be large enough to invert a ranking.

Wet lease and capacity purchase. When an ACMI operator flies the sector, attribution depends on the feed’s convention: the contracting carrier, the operating certificate holder, or both. Compare two sources with different conventions and the same market appears to grow or shrink for reasons that have nothing to do with aeroplanes.

Seats that exist but are not sellable. Blocked seats for crew rest or positioning, seats withheld for weight and balance on performance-limited sectors, cabins partly closed for maintenance or for cargo work. These never appear in a schedule file, and only sometimes in a traffic return.

Airlines plan flights with a certain number of seats, but many things can reduce the actual seats flown. Flights get canceled, smaller planes are used, or some seats aren’t available for passengers. The numbers you see in reports often don’t match what really happens.

An airline files a schedule for a daily A320 flight from Paris to Madrid with 180 seats. Due to a technical issue, the flight is downgauged to an A319 with 140 seats on some days. Additionally, 10 seats are blocked for operational reasons (inoperative seats, weight and balance). The actual flown seats are now 130 instead of the planned 180.

Where seats disappear between the filed schedule and the flown programme
Mechanism What it removes Departure survives? Visible in a schedule file?
Cancellation Whole flight: seats, ASK, movement No Only if the file is refreshed after the fact
Equipment substitution Part of the seat count Yes No — the filed type is what stays on the record
Configuration default Accuracy of every seat figure Yes No — the type code hides the cabin
Marketing duplication Nothing — it adds phantom seats Yes Yes, if operating and marketing rows are flagged
Wet lease attribution Nothing — it moves seats between carriers Yes Depends entirely on the feed’s convention
Blocked or unsellable seats Sellable capacity only Yes No

The snapshot date changes the answer

A schedule file has no single value. It has a value as at the moment it was pulled, and it decays as the season approaches. Pull a programme six months out and it contains flights that will be trimmed, times that will move and aircraft that will change. Pull it the day before departure and it is close to reality — and useless for planning.

Part of that decay is ordinary commercial adjustment. Part of it is structural to how slots work. At coordinated airports, capacity is allocated through the seasonal process described in the industry’s worldwide slot guidelines and, in the European Union, additionally governed by the European slot regulation. Historic precedence depends on a usage threshold — the familiar use-it-or-lose-it principle, set at eighty per cent of the allocated series under the European regime in normal times, and repeatedly relaxed by emergency amendment during the pandemic years. An airline that files a series to protect precedence and later reduces it has done nothing unusual, but its early filing was never a forecast of flying.

The consequence for analysis is blunt. Comparing a schedule pulled at ninety days against one pulled at thirty days is not a measurement of capacity change; it is a measurement of schedule maturity. Any year-on-year comparison has to hold the lead time constant, or it compares two different objects and attributes the difference to the market.

The trap. A « capacity down eleven per cent » headline built from a current-season pull against a completed prior season compares a maturing plan with a settled record. The plan will keep shrinking; the record cannot. The comparison is biased before a single flight is examined.

Seats, departures and ASK do not shrink together

The three usual metrics respond differently to the same event, which is why they have to be chosen against the question rather than quoted together as if interchangeable.

Cancel a regional turboprop rotation and you lose one departure and a small number of seats. Cancel a long-haul widebody and you lose one departure, a large number of seats and a very large number of available seat kilometres. Downgauge a narrowbody and you lose seats while the departure count is untouched. Reroute a long sector around closed airspace and the aeroplane burns considerably more fuel over considerably more track — but great-circle-based available seat kilometres will not notice, which is a distortion running in the opposite direction to all the others.

Choosing the capacity metric against the question being asked
Metric Question it answers Blind to Typical misuse
Departures and movements Airport and airspace workload, slot pressure, ground handling demand Aircraft size entirely Quoted as « capacity » in a passenger market
Seats Offer to a specific market, competitive share on a route Distance, and therefore cost and revenue weight Summed across marketing carriers, double-counting code-shares
Available seat kilometres System-wide production, unit cost denominators Route-level competitive reality, and the track actually flown Compared across carriers with very different network shapes
Sellable seats Commercial capacity genuinely offered Almost never published at all Assumed identical to installed seats

What a flown source looks like

The corrective is a source built after the fact, from operations rather than from intention. Several exist. None is global, all are lagged, and each counts something slightly different.

The cleanest structural example is the segment traffic reporting required of United States carriers, where a single record carries departures scheduled and departures performed alongside seats and passengers, broken out by carrier, segment and aircraft type. That pairing is exactly the ratio a schedule file cannot produce. For anyone working a North American or transatlantic question it is the natural anchor, and its structure is worth studying even where it does not apply, because it shows what a complete capacity record needs to contain.

In Europe, network flight counts from the air navigation side answer movement questions with very little lag, but they count flights, not seats. Airport authorities publish movements and passengers, again with no seat dimension. Airlines publish monthly traffic — available seat kilometres, revenue passenger kilometres, load factor — on their own definitions, which differ on charter flying, on subsidiary consolidation and on the treatment of wet-leased capacity. States report traffic to the international statistics programme through the standard air transport reporting forms, at an aggregation and a lag that suit macroeconomic work rather than route decisions.

The practical position is that no single feed answers a route question end to end. The usable technique is calibration: find one period where a schedule pull and a flown return both exist for the same market, compute the realisation ratio between them, and carry that ratio forward as an explicit, written assumption rather than an invisible one.

Load factor is a third quantity again

Once capacity is corrected, a second confusion is waiting. Load factor is passengers carried divided by seats flown, not seats filed. A carrier that cancels aggressively through a weak month will report an excellent load factor on a shrunken denominator, and the figure will be entirely true and entirely misleading about demand.

The arithmetic works the other way too. A carrier that protects its programme through a soft period absorbs the softness in load factor rather than in capacity, and looks weaker on the headline ratio while having carried more people and, quite possibly, more revenue. Any read on demand needs the numerator in front of it — passengers, or revenue passenger kilometres — not the ratio on its own.

The right reflex. Before quoting a load factor movement, write down what the denominator did. If capacity moved more than the ratio did, the story is a capacity story and the load factor is a by-product of it.

Correcting the read before it becomes a decision

None of this is an argument against schedule data. It is the only forward-looking source in existence, it is timely, and for competitive monitoring nothing replaces it. It simply needs a small number of corrections applied in the open, so that whoever reads the conclusion can see the assumptions that produced it.

Four corrections to apply to any schedule-derived capacity figure
Correction How it is done What it fixes What it looks like when skipped
Collapse to the operating flight Filter on the operating carrier flag and drop marketing duplicates Phantom seats on code-shared markets A market shows more seats than its movement count can physically carry
Replace default seat counts Map registration or sub-fleet to the operator’s published cabin Type-code averaging across mixed configurations Leisure and long-haul variants of one type valued identically
Apply a realisation ratio Calibrate the schedule against a flown return for a comparable season and market Cancellations and equipment substitution Consistent optimism, largest in the most disrupted seasons
Fix the snapshot lead time Compare pulls taken the same number of days before operation Schedule maturity masquerading as growth or decline Year-on-year swings that quietly reverse as the season matures

After the corrections, one further step separates analysis from decoration: test whether the decision survives the residual uncertainty. If a route case is positive on filed capacity and negative once a plausible realisation ratio is applied, the case is not marginal — it is unproven, and the honest output is a range with the assumption named beside it. Practitioners comparing methods across markets usually find the reconciliation argument more instructive than the datasets themselves, and the FL330 forums carry a steady traffic of exactly that kind of work.

The sentence that belongs beside every capacity number

A capacity figure without provenance is not a measurement. Three facts turn it into one: the source and its snapshot date, whether it counts filed or flown flights, and whether marketing duplicates were removed. Written out, it reads something like « seats, operating carrier only, schedule pull at sixty days before operation, uncorrected for cancellations ». That is a sentence a colleague can argue with, which is the whole point.

The alternative — a bare percentage with a season attached — invites the reader to assume it is a record of what happened. It rarely is. The published figure is the plan, and the plan sits at the optimistic end of the distribution, because the mechanisms that erode it are numerous and the mechanisms that inflate it are two: extra sections added late, and counting the same aeroplane twice.

Commentaires

Leave a Reply

Découvrez l’écosystème FL330

FL330, c’est toute l’aviation professionnelle — explorez nos autres espaces.