Discretionary Fuel: What the Drift in the Uplift Actually Measures

A pilot in the cockpit reviewing a flight plan and fuel calculations at dusk, with runway lights visible outside

Every quantity on a flight plan is computed by someone other than the crew — except one. Discretionary fuel is the commander’s own number, and because it is the only free variable in the whole document, it is also the only place where the crew’s opinion of the plan becomes measurable. Read one uplift and you learn nothing. Read a year of them, sorted properly, and you are looking at a survey of how much your own planning system is trusted.

The only free number in the fuel plan

The fuel order that comes out of a planning system is a stack of named quantities, each answering one question and no other. Taxi fuel answers ground movement. Trip fuel answers the flight from take-off to landing at destination. Contingency fuel answers deviation from the predicted trip consumption. Alternate fuel answers the missed approach and the transit to an alternate aerodrome. Final reserve answers the last, inviolable holding period, and additional fuel answers a specific identified requirement such as an isolated aerodrome or a driftdown case.

Every one of those is derived. A computer applied a rule to an input, and if you disagree with the answer, the honest response is to argue about the rule or the input. None of them is a judgement call at the aircraft.

Then there is the last line: the fuel the commander adds because, having looked at the whole picture, they want more. European and ICAO frameworks both leave that authority explicitly with the commander, who may add fuel but may never depart with less than the minimum the applicable scheme requires. It has several names in several manuals — discretionary fuel, extra fuel, captain’s fuel — and one defining property. Nothing computed it. Somebody decided it.

That property is what makes it an instrument. A derived number tells you about the model. A decided number tells you about the person deciding, and about what they believe the model is getting wrong.

What discretionary fuel is not

Three confusions do most of the damage in fuel discussions, and they are worth clearing before any data is looked at.

It is not contingency fuel. Contingency exists to absorb the difference between predicted and actual trip consumption — longer routing, a lower level, wind that did not verify, an airframe burning above the model. It has a defined basis and, under a statistical scheme, a defensible one. If a crew is systematically adding discretionary fuel to cover consumption deviation, the contingency basis is wrong and the discretionary uplift is a symptom, not a solution.

It is not additional fuel. Additional fuel covers identified, nameable requirements that the planning process is supposed to catch: an isolated destination with no usable alternate, a driftdown or depressurisation case, a fuel-relevant item on the minimum equipment list. If those keep arriving as discretionary fuel instead, the identification step has failed upstream.

And it is not a safety margin on top of the final reserve. The final reserve is not a target and not a buffer to be eaten into; planning to arrive below it is not permitted, and in flight, a calculated landing fuel below final reserve at the nearest aerodrome where a safe landing can be made is an emergency declaration, not a discussion. Discretionary fuel does not make the final reserve larger. It buys options — a longer hold, a second approach, the ability to accept a delay rather than divert — and options are worth paying for when they are the ones actually needed.

The five questions a fuel plan answers, and which one the commander owns
Quantity Answers Set by Signal value when it moves
Trip fuel The planned flight itself Performance model and route Model accuracy, aircraft condition
Contingency fuel Deviation from predicted consumption Rule or statistical scheme Data quality of the scheme
Alternate and final reserve Diversion and the last holding period Regulation and route None — they are floors
Additional fuel An identified specific requirement Planning, case by case Whether identification is working
Discretionary fuel Whatever the commander does not trust The commander alone Confidence in everything above

The carriage cost, and why it is the weaker argument

Carrying fuel costs fuel. The extra mass has to be lifted, accelerated and held aloft, and a fraction of it is burnt simply to be there. The fraction is aircraft-specific and sector-specific; on most transport jets it amounts to a low single-digit percentage of the carried mass for each hour aloft, and the performance department can give the figure for each type. Over a fleet and a year the totals are large enough to make finance departments interested, and that interest is where most discretionary-fuel programmes begin — and where a good many of them go wrong.

The trouble with leading on cost is that it puts the operator in the position of asking crews to carry less of something whose whole purpose is to be there when the plan fails. That is an argument the operator cannot win in the crew room, should not win, and does not need to win. The stronger argument is different and entirely compatible with the commander’s authority: if crews are consistently adding fuel on a given route, either the plan for that route is wrong or the crews believe something the operator has not measured. In both cases, the operator has a problem that carrying less fuel does not solve.

There are also physical costs that have nothing to do with money. Extra fuel is extra landing weight, which matters where landing performance is limiting, where brake energy is a constraint on a quick turnaround, or where a maximum-landing-weight arrival forces a decision nobody wanted. Those constraints are worth naming explicitly, because they are the ones a crew will accept without feeling that safety is being priced.

Discretionary fuel is the extra fuel pilots add because they don’t fully trust the flight plan. It’s not about safety margins or emergencies—it’s about covering gaps in the plan that the pilots notice. If many pilots keep adding fuel for the same reasons, it means the planning system needs fixing, not the pilots.

On a busy summer day, an Airbus A320 flying from Paris to Nice consistently has crews adding 1,000 kg of discretionary fuel. Analysis shows that this is because the flight plan doesn’t account for predictable holding at Nice due to traffic congestion. The operator adjusts the plan, and the extra fuel drops to zero—proving the system was the issue, not the pilots.

Reading the drift: four patterns, four different diagnoses

A single discretionary uplift is unreadable. The signal appears only when the uplifts are sorted, and the sorting that matters is two-dimensional: does the extra fuel cluster on a route, or on a person?

If it clusters on a route and is spread evenly across commanders, the crews are agreeing with one another about something. That is not a behaviour problem; it is a plan problem, and it is the most valuable finding a fuel programme can produce. Independent professionals converging on the same correction almost always means the reference computation is missing something real.

If it clusters on individuals and is spread evenly across routes, the plan is probably sound and the variation is personal — experience, an unresolved past event, or simply never having been shown what the numbers do. That is a training and briefing matter, and it is handled individually and privately, never through a league table.

If it clusters on a tail number, look at the performance factor before looking at anything else. And if it is spread evenly across everything, the operator is looking at a general level of trust in the planning system rather than at any specific defect — a slow, chronic reading that moves only when the system visibly improves.

Where the extra fuel clusters, and what it is telling you
Pattern Most likely cause Where to look first Who owns the fix
By route, across all commanders The reference plan for that route is wrong Filed versus flown track, level capping, arrival holding history Flight planning
By individual, across all routes Personal calibration, experience or an unresolved event A conversation, not a report Fleet management, privately
By tail number Stale or wrong aircraft performance factor Cruise performance monitoring and its update cycle Engineering and performance
By season or time of day Weather, flow control or a known congestion window Delay and holding statistics for that period Network and planning together

When the plan is the problem

Three planning defects produce discretionary uplift with remarkable reliability, and all three are measurable.

The first is the aircraft performance factor. Planning systems carry a per-airframe correction that expresses how far the real aeroplane is from the book, and it is maintained from in-flight cruise performance monitoring. When that factor is stale — after an engine change, after a period without monitoring, after a modification — the plan under-predicts the burn, the crews notice within a handful of sectors, and they compensate the only way available to them. Discretionary fuel that follows a tail number rather than a route is very nearly a diagnostic for it.

The second is route realism. A plan computed on a route that air traffic control never actually assigns is a plan that is wrong before the aircraft moves. So is one computed at a level routinely unavailable in that flow. Crews learn these things sector by sector, long before any report captures them, and the comparison of filed track against flown track is one of the cheapest analyses an operator can run.

The third is arrival reality. If a destination reliably holds at a particular hour, that holding is neither a contingency nor a discretionary matter — it is a known, anticipatable requirement, and it belongs in the computed plan under whatever heading the operator’s scheme provides. Leaving it to the commander to rediscover every day converts an operational fact into a personal judgement, and then invites complaints about the variance in the judgements.

When the forecast is the problem

The other half of the trust equation is meteorology, and it splits into two very different questions.

The first is upper-wind quality. The wind field used by the planning system has an issue time and a validity, and a plan built on an ageing field will drift in a direction that is often systematic rather than random on a given route or in a given season. Verifying forecast wind against recorded actual wind, aggregated by route, tells you whether the crews’ scepticism is justified. Often it is.

The second is terminal forecast quality, and it behaves quite differently. What drives discretionary uplift here is not the forecast being wrong but the forecast being uncertain — a temporary or probabilistic group that straddles a minimum, a trend that could go either way, a single-runway destination with a marginal crosswind. Crews do not add fuel because they expect the bad case; they add it because the plan gives them no way of buying the option to wait. When an operator can show that its own record of such days rarely required the diversion, the conversation becomes possible. When it cannot show that, it has no standing in the conversation at all.

The trap. The moment discretionary uplift is published per commander, ranked, or attached to any form of personal assessment, the instrument stops working. Crews do not become less cautious; they become less legible. Some take the fuel and code the reason vaguely, some take it and stop coding at all, and a few take less than they wanted, which is the outcome nobody claims to want and everybody risks producing. A number that is used to judge people cannot also be used to measure a system.

Reason codes that survive contact with a crew room

For the drift to be readable, the reason has to be recorded, and recording it has to cost almost nothing. That constrains the design more than most operators expect.

  • A short list works; a long one does not. Six to eight codes, chosen so that a crew member can find the right one in a couple of seconds, will produce usable data. Twenty codes will produce whichever one is at the top of the list. A free-text field alone produces beautiful anecdotes and no statistics; a free-text field alongside a short code list produces both.
  • The codes should describe the world, not the person. « Expected holding at destination », « forecast uncertainty at destination », « expected routing longer than filed », « expected level capping », « aircraft burning above plan », « de-icing or ground delay expected », « crew judgement, see remarks ». Note that the last one is not a failure of the scheme — it is the escape hatch that keeps the other seven honest, and its frequency is itself a measurement.

The right reflex. Close the loop out loud. Publish, at fleet level and never at individual level, what the extra fuel bought over the last quarter: how often it was consumed, how many diversions it prevented, how many holdings it absorbed. Crews are entirely willing to adjust a habit when shown the outcome of the habit. What they will not do — and should not do — is adjust it because a spreadsheet said the fleet burnt a percentage too much. The fuel policy modules on the FL330 training site take the same line: the analysis exists to correct the plan, not to grade the pilot.

The line the operator cannot cross

Everything above is an argument for measuring discretionary fuel carefully. None of it is an argument for constraining it, and the distinction has to be maintained explicitly, in writing, in the operations manual.

A man in a uniform uses a tablet to discuss data with two colleagues in a cockpit at dusk.
Data-driven decisions optimize fuel efficiency in real time.

The commander’s authority to add fuel is not a courtesy granted by the operator. It exists because the person on the aircraft has information the planning system does not — the actual weather they can see, the actual state of the machine, the actual flow situation, and their own assessment of the day. An operator that reduces its average uplift by making that authority uncomfortable to exercise has not improved anything. It has removed its own last line of defence against a planning error and, worse, it has removed the instrument that would have told it the error existed.

The productive version of the programme is the opposite in tone and identical in arithmetic: make the plan good enough that the extra fuel becomes unnecessary, and let the falling uplift be the evidence that it worked. That figure descends on its own when the reference computation earns trust. It is exceptionally durable when it does, because nobody was ever asked to carry less than they judged necessary — the judgement simply changed.

Commentaires

Leave a Reply

Découvrez l’écosystème FL330

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