Industry: Aerospace Product: ModelRisk Application: Retrofit and Upgrade ROI Under Uncertainty
A long-haul operator was evaluating a fleet-wide retrofit of 62 wide-bodies: new IFE and connectivity (≈$4.8M/tail), winglet retrofit for fuel savings (≈$3.6M/tail), and cabin reconfiguration to add 14 premium-economy seats per aircraft (≈$5.6M/tail). Total program $868M at central cost estimates. The deterministic NPV showed a single $312M lifetime NPV at an 8.5% discount rate — a modest return on an $868M outlay, and silent on how much could go wrong.
The Monte Carlo program model told a richer story. Run across 25,000 twelve-year paths, the full retrofit returned a mean NPV of $772M with a P10 of –$519M, a P50 of $425M, and a P90 of $2,488M — and a 32% probability of loss. The full program is attractive in expectation, but it is a genuine risk: roughly one path in three loses money, because the premium-economy revenue, the fuel savings it monetises, and the residual uplift all ride on one shared travel-demand and fuel-price cycle that can turn down across the whole fleet at once. The deterministic $312M sat well inside the distribution but hid both the upside and that downside. The investment committee approved the full program on the strength of that distribution — and explicitly rejected the partial scopes the deterministic framing had favored, because they lose money outright.
The three histograms tell the whole decision at a glance: the full program (red) has a mean of $772M and most of its mass to the right of zero — though a real left tail crosses below it — while both descoped options pile up entirely to the left of zero. Stripping out the cabin reconfiguration does not de-risk the program — it removes the premium-economy revenue stream that carries the return, leaving an investment that loses money with near-certainty.
Fuel savings from winglet retrofit — supplier-quoted 3.2% block-fuel reduction, but the real-world realized number depended on flight-mix, weight management, and engine condition. Modeled as Beta on [0.018, 0.046] with mean 0.030 — bounded around the supplier figure because the upper-bound promise rarely held up in long-haul operation. Per-aircraft annual fuel burn was 6.8M lb; jet fuel price modeled as the operator's documented two-regime jump-diffusion (low-vol drift 2%/yr, σ 26%; crisis drift 5%/yr, σ 52%, switching probability 11%/year).
Premium-economy revenue uplift — the single biggest swing in the program. 14 incremental Y+ seats per aircraft at a yield uplift modeled as LogNormal with median $820 per seat per leg above the displaced Y seat (net of cannibalisation), σ_log 0.50, and then scaled by a shared market-cycle factor common to the whole fleet. Premium-economy load factor was Beta(5.0, 3.5) on [0.40, 0.92], itself tilted by the same market cycle, reflecting the slower ramp documented in published comparable cabin reconfigurations. The seats had to be filled before they paid back the cabin spend.
Maintenance cost reduction from new IFE — modeled as LogNormal with median 30% reduction in IFE-related MMH/FH and σ_log 0.45. (The prior IFE generation had a documented MTBF of 410 flight hours per seat; the new system was supplier-rated at 2,800 FH, but realised reliability — and so the realised saving — was uncertain and widened accordingly.)
Downtime cost during retrofit — each aircraft loses an average of 18 days to the retrofit (Triangular 14 / 18 / 30 days), at a lost-revenue cost of LogNormal median $360k/day, σ_log 0.40. The program induces tens of millions of mean downtime cost on the way to the benefits.
Residual-value uplift — a retrofitted aircraft's year-12 sale value benefits from a documented secondary-market premium for premium-economy-equipped, winglet-equipped, connectivity-equipped wide-bodies. Modeled as a 1%–8% uplift on residual value, positively correlated (ρ=0.50, Gaussian copula) to the fuel-price regime — a high-fuel environment makes fuel-efficient retrofits worth more on the secondary market.
25,000 12-year simulation paths discounted at 8.5% produced the program NPV distribution shown above. For the full program: mean NPV $772M, P10 –$519M, P50 $425M, P90 $2,488M, P(NPV < 0) = 32%. The deterministic estimate at central inputs sat at $312M — below the simulated mean because it freezes the premium-economy yield and load factor at their medians and captures none of the right-skewed upside, but also blind to the left tail. A single shared market-cycle common factor drives every aircraft's premium-economy revenue and fuel saving together, so the 62 tails do not diversify the risk away: a weak travel-and-fuel cycle pulls the whole fleet's benefit down at once, which is what produces the genuine one-in-three chance of loss rather than a central-limit collapse to a risk-free mean.
The model was rerun under three program scopes to test whether a partial retrofit dominated:
The premium-economy yield distribution drives the largest spread by far (≈±$1,306M half-spread, a $2,612M P10–P90 range) — the cabin spend's NPV is hostage to a right-skewed yield with a long tail. The shared market-cycle common factor is second (≈±$817M), confirming that the program's risk is systematic, not diversifiable across the 62 tails. The premium-economy load factor is third (≈±$688M). Retrofit downtime cost, the fuel-price regime, and the residual-uplift × fuel-regime copula follow well behind. The winglet block-fuel-savings parameter is bounded and contributes almost nothing to the spread — the savings are small relative to the cabin revenue that drives the program.
A single $312M deterministic NPV hid that the cabin reconfiguration was the engine of the return, not a variance-heavy luxury — and that the right corporate answer was to fund the full program rather than the "safe" partial scope, which the simulation showed loses money with near-certainty.