Collect independent signals
We use publicly available information and independent fleet observations collected directly or on our behalf by field researchers.
Evidence before certainty
How our forecast works
We combine verified fleet observations and traveler evidence into a likelihood score for each model in the rental category you booked, while showing separately how mature the underlying evidence is.
Verify identity and category separately
A vehicle model must be visually verified. Its rental category must come from the rental company's official terminology, so we do not guess class from body shape.
Weight each evidence type differently
Fleet scans describe the composition of a lot, reviewed ActualFleet pickup reports confirm what a traveler received, and public reports add supporting evidence. One large scan is deliberately prevented from behaving like dozens of independent reports.
Let older evidence fade gradually
Traveler reports lose half of their weight for each year of age instead of disappearing at a hard cutoff. Scan evidence is also reduced as it gets older.
Turn evidence into a 0–100 score
Each model receives a likelihood score based on relative support, the strength of the evidence behind that model, and evidence quality and freshness. The score is a ranking signal, not a percentage probability.
Likelihood score
A useful number without pretending it is a probability
The 0–100 score measures how strongly the available evidence points to one model relative to the alternatives in the same airport, rental company, and booked category. A score of 72 does not mean a 72% chance of receiving that model.
Evidence maturity
Early, growing, or strong evidence
Evidence maturity is shown separately from the model score. It increases as we collect more independent verified fleet checks, more observation dates, more verified vehicles, and more reviewed ActualFleet pickup reports.
Current v2 score
Relative support 60%, strength 25%, quality 15%
Relative support measures how dominant a model is among the evidence we have, with sparse categories deliberately pulled toward the middle. Model evidence strength rewards repeated support. Evidence quality favors verified scans and reviewed ActualFleet reports over public reports, and all evidence is reduced as it ages.
Important limitation
Observed pattern, not guaranteed inventory
Airport fleets move continuously. Likelihood scores are intended to help set expectations and compare options, but the exact vehicle at pickup will always depend on live availability.
MCO pilot
View the Standard SUV forecast →