Methodology

1. Source vs. derived data

Every price you see on Empty Leg Index falls into one of two categories.

Source data

The listed price, aircraft type, origin, destination, and departure time for every leg come directly from operator and broker listings. We aggregate across multiple sources, refresh every 15 minutes, and never modify the underlying values. When a listing's price changes, we record a snapshot in our price-history table — the raw material for the decay analysis below.

Derived data

Three numbers per listing are our estimates, computed using documented formulas and published reference tables:

• The estimated operator floor (section 2)
• The price-decay context for the route class (section 3)
• The wait-or-book recommendation derived from the two above (section 4)

We label these clearly throughout the product and never present them as quoted prices. Our estimates can be wrong — we publish the math so anyone, operators included, can challenge it.

2. The fuel-floor estimator

Concept

When an operator has committed to flying an aircraft on a one-way charter, the return leg is a sunk cost. Crew is paid; the airframe will move. Any revenue on that leg is incremental. The operator's true minimum is the marginal cost of operating it: fuel, crew duty time, and ground handling at both ends.

We call this the operator floor. The estimate is conservative — a price the operator could rationally accept if no other revenue is available, not what they're targeting in negotiation.

The formula

estimated_floor = fuel_cost + crew_cost + ground_handling + positioning_buffer

where:
  block_time_hours      = (great_circle_distance / cruise_kts) × 1.10
  fuel_burn_gal         = block_time_hours × cruise_gph + reserve_gal
  fuel_cost             = fuel_burn_gal × jet_a_price_per_gal
  crew_cost             = aircraft.min_crew_cost_per_leg
  ground_handling       = origin_handling + landing_origin
                        + destination_handling + landing_destination
  positioning_buffer    = if distance < 200nm: 0.10 × subtotal else 0

Multiplier rationale

• 1.10 on block time — taxi, climb, descent, and ATC routing add roughly 8–12% over great-circle time. 1.10 is the median in published charter-quote data.
• Reserve fuel — 100 gal for light, scaling to 350 gal for ultra-long-range. Conservative interpretation of NBAA IFR reserves.
• Crew cost — represents marginal duty-time cost for a leg, not fully-loaded crew salary.
• Positioning buffer — under ~200 nm, fixed costs dominate. Adds 10% to the otherwise-calculated subtotal.

Display range and confidence

We display the floor as a range — floor × 0.92 to floor × 1.08 — to communicate uncertainty honestly.

• High — all reference data was direct. No fallbacks.
• Medium — exactly one fallback (e.g., FBO category default).
• Low — multiple fallbacks, OR distance < 100 nm. Treat as ballpark; we suppress recommendations in this case.

Reference data — aircraft (0 models)

Cruise GPH and KTS values are conservative interpretations of book performance figures; real-world burns are typically 5–10% higher.

Download as CSV →

Reference data — airports (0 airports)

Costs are typical-range estimates from published FBO rate cards, cross-referenced with operator-reported figures. Airports without explicit data fall through to category averages.

Download as CSV →

Jet-A fuel price

We currently use a single national-average Jet-A price, refreshed weekly. As of 2026-04-27, we're using $6.25/gallon.(stale — refresh pending) Regional pricing refinement is on the roadmap.

3. The price-decay curve

Empty leg prices typically soften as the departure date approaches. The closer to departure, the less time the operator has to find another buyer, and the more willing they are to settle near floor. We measure this directly.

Method

hours_to_departure = listing.departure_at - observation.observed_at
price_index        = observation.price / listing.first_observed_price

route_class        = aircraft_category × distance_band

distance bands:    short (<500 nm), medium (500-1500 nm), long (>1500 nm)
buckets:           >168h, 72-168h, 48-72h, 24-48h, 12-24h, <12h

Sample-size threshold

We only display a curve when the route class has at least 50 unique completed listings in the last 12 months. Below that threshold, the median is too noisy.

Outlier handling

We exclude observations where price_index < 0.4 or price_index > 1.5 — these are typically data errors. Negative hours_to_departure values past 2h are excluded as scrape-timing errors.

Current curves (0 route classes, 0 listings analyzed)

No route classes have sufficient sample size yet. Curves appear once we accumulate 50+ completed listings per class.

See the full decay report →

4. Wait-or-book recommendations

Given a listing's floor estimate, the historical decay curve for its route class, and time remaining to departure, we produce one of five recommendations: BOOK NOW, WAIT, WAIT or NEGOTIATE, FAIR, or INSUFFICIENT DATA.

Decision tree

1. If floor confidence is low, return INSUFFICIENT DATA.
2. If listed price is at or below estimated floor, return BOOK NOW with note “exceptional deal.”
3. If < 6 hours to departure:
   • Listed within 25% of floor → BOOK NOW
   • Else → WAIT or NEGOTIATE (operators rarely refuse cash offers in this window)
4. If listed price is within 15% of floor → BOOK NOW (“already strong price”).
5. If we don't have a usable decay curve for this route class:
   • Spread < 30% above floor → FAIR
   • Spread ≥ 30% above floor → WAIT
6. Compute expected remaining decay from the curve. If expected final spread is at least 10 percentage points below current spread → WAIT. Otherwise → FAIR.

What “BOOK NOW” means

It does not mean we predict you'll save money.It means: this listing is already at or near the price we'd expect it to settle at, given current market conditions for its route class. Further drops are unlikely to be meaningful, and waiting risks losing the deal. A recommendation, not a guarantee.

What “WAIT” means

Listings in this route class typically drop in price between now and departure. Based on historical medians, you can expect a meaningfully lower price closer to the date. It does not mean a specific listing will drop — only that the central tendency for legs like this is to soften.

Sellout-risk numbers

The Founding tier displays an estimated risk of the listing selling before the typical decay window elapses. v1 uses static estimates by departure-time bucket (3% at >7 days out, scaling to 30% under 12 hours). v1.1 will replace these with empirical rates measured against booking outcomes once we have ≥3 months of recommendation-served data.

Calibration tracking

Every recommendation we serve is logged with its inputs and timestamp. We periodically measure how often WAIT recommendations were followed by actual price drops (and BOOK NOW recommendations were followed by listings being booked). When we update the decision rules based on this feedback, the change appears in the changelog.

5. What we don't claim

The honest list of where our methodology is limited:

• We don't predict individual listings. Our recommendations describe central tendencies; the specific listing may behave differently.
• We don't model operator-specific economics. Newer fleets, owner-operators, and partner deadhead deals can run leaner. Our floor is an industry-average baseline.
• Our floor is an estimate, not a quote. Actual willingness to sell depends on factors we don't observe (other inventory, scheduling, customer relationships).
• We don't account for special services. Catering, specific FBO selection, alternate routing — none of this is in the floor.
• The decay-curve method assumes the past predicts the present. A market shock would invalidate near-term curves until we accumulate new data.
• We don't model sales-tax differences across jurisdictions. Listed prices may or may not include tax; we display them as-is.

Operator corrections welcome

We get reference data wrong sometimes. If you're an operator and you see a number that's off — handling fees at a specific FBO, fuel-burn rates for an aircraft, anything — email [email protected] with your correction. We'll acknowledge within 48 hours, investigate, and credit you in the changelog.

6. Methodology changelog

Every change to a formula, multiplier, or reference dataset is logged here. We also log dataset refreshes (e.g., quarterly aircraft profile review).