From observed fares to an auditable index.
The frozen framework defines comparable fares and separate indices for each booking horizon. The current prototype implements the core calculations on demo data; proposed production features and official validation data are identified below.
One series per advance-booking window
Booking window = calendar date of travel − calendar date of collection. Observations from different horizons are never combined in an elementary route index.
DATA COLLECTION
Source adapters retain the raw observation, source details, collection time and declared data mode. This demo uses synthetic fixtures; it does not perform live fare scraping.
FARE NORMALIZATION
Comparable fare = Base Fare + UDF + ASF + GST + YQ. Convenience fees and optional extras are excluded; a total-minus-convenience-fee fallback is used when the component breakdown is incomplete.
QUALITY CONTROL
The prototype checks component consistency, screens commercial duplicates and classifies price anomalies. Invalid component totals and technical outliers are excluded from index eligibility.
JEVONS ROUTE INDEX
For each route and booking horizon, matched flight price relatives are combined as a geometric mean against the persisted base period.
NATIONAL AGGREGATION
Route indices are combined with passenger-volume reference weights using a weighted arithmetic (Young / Modified Laspeyres) or geometric (Jevons) mean.
VALIDATION
Calculated values are date-aligned with a configured reference series. Metrics require matched observations; they are not evidence of official benchmark accuracy.
PROVENANCE
Every parsed observation retains a trace to its raw payload and SHA-256 checksum.
Index formulas
- Comparable fare: base fare + UDF + ASF + GST + mandatory YQ surcharge.
- Route × horizon (Jevons): 100 × exp(mean[ln(current fare ÷ base fare)]).
- National Young / Modified Laspeyres: sum(route index × route weight) ÷ sum(observed route weights).
- Directional markets: A→B and B→A are separate route series.
- Missing fares: remain missing; no index-value imputation is applied.
Comparability & anomaly rules
- Excluded from fare: convenience fees, optional baggage, seat selection, meals, insurance and conditional promotions.
- Component check: a full breakdown must reconcile to the displayed total (with or without convenience fee); mismatches are disqualified.
- Deduplication: matches origin/destination, airline, flight, travel date, departure time, horizon, cabin and fare family; direct airline sources outrank OTAs, then the lowest fare wins within a tier.
- Tukey screen: the prototype uses a 1.5 × IQR upper fence; elevated fares below 3.5 × the cell median remain possible market surges, not automatic exclusions.
- Small cells: outlier screening is skipped when fewer than four positive fares are available.
Weights, validation & auditability
The framework specifies DGCA passenger-volume reference weights. Values bundled with this synthetic demo are not official DGCA weights. Validation compares matched dates and reports MAPE, RMSE, bias, correlation and directional accuracy when available; the prototype can emit descriptive metrics from three matched pairs, while the research validation guardrail calls for at least 15 pairs before making a confidence claim. Provenance preserves LIVE, HISTORICAL or SYNTHETIC mode and traces records back to their raw payload hash.
Scope & disclosure
Synthetic composite-horizon weights, imputation and automated live scraping are not presented as active methodology here. The reference baseline is synthetic demo data—not official DGCA or MoSPI airfare microdata—and prototype metrics do not establish production accuracy.