NP-RM-08 · RM System Calibration & Tuning

📊

Process Flow Diagram (BPMN)

📋

L4 Process Steps

Step	Step Name	Role / Swim Lane	System	Input	Output	KPI	Dec?	Exc?
Phase 1 1.1	Extract RM forecast accuracy metrics	Revenue Management Analyst	Amadeus Revenue Management (NRM)	Scheduled monthly calibration cycle or ad-hoc trigger from revenue variance alert	Forecast accuracy report: MAPE by market, booking class, and horizon	Mean Absolute Percentage Error (MAPE) ≤15% across all O&D markets	N	N
1.2	Identify underperforming O&D markets	Senior Revenue Analyst	Tableau	Forecast accuracy report from step 1.1; actuals from AWS Redshift booking data warehouse	Ranked list of O&D markets with MAPE >15% or revenue variance >3% vs. plan	≥90% of O&D markets within MAPE threshold within 30 days of calibration cycle	N	N
1.3	Assess booking curve deviations from expected pickup	Revenue Optimization Analyst	Amadeus Revenue Management (NRM)	Ranked O&D market list; historical booking curve profiles from NRM	Deviation analysis report flagging systematic early/late booking shifts and class-mix anomalies	Booking curve R² ≥0.92 across flagged markets; decision to proceed with recalibration if ≥3 markets fail threshold	Y	N
Phase 2 2.1	Extract historical booking and ticketing data	Data Engineer	AWS Redshift	Calibration scope (markets, date range, booking classes) from deviation analysis report	Structured training dataset: O&D bookings, fares paid, cancellation rates, no-show rates — 24-month rolling window	Dataset completeness ≥98% of scheduled departures in scope; extract runtime <2 hours	N	N
2.2	Validate input data quality and completeness	Data Quality Analyst	AWS S3	Raw extract from AWS Redshift; data quality ruleset stored in S3	Data quality scorecard with pass/fail per validation rule; cleansed dataset or rejection notice	Data quality score ≥95% (null rate <2%, duplicate bookings <0.5%, fare outliers <1%)	Y	Y
2.3	Load competitive fare and capacity benchmarks	Revenue Management Analyst	ATPCO	Competitor fare filing data from ATPCO; OAG Schedule Analyser capacity data for target O&D markets	Competitive context dataset: competitor fare ladder, capacity by carrier, schedule frequency	Competitive data coverage ≥85% of target O&D markets; data freshness within 48 hours of filing date	N	N
Phase 3 3.1	Recalibrate demand forecast model coefficients	RM Systems Analyst	Amadeus Revenue Management (NRM)	Cleansed 24-month historical dataset; competitive context dataset; current NRM model parameter set	Updated demand model coefficients (seasonality indices, trend factors, day-of-week multipliers) staged in NRM test partition	Post-recalibration MAPE improvement ≥10% vs. baseline on hold-out validation set	N	N
3.2	Update price elasticity and WTP model curves	Revenue Optimization Analyst	Amadeus Revenue Management (NRM)	Updated demand model coefficients; ATPCO competitive fare ladder; passenger segment willingness-to-pay data	Revised price elasticity curves and class-specific bid price adjustments per O&D market cluster	Bid price accuracy: simulated revenue within ±2% of optimal on hold-out markets; elasticity curve R² ≥0.88	N	N
3.3	Tune overbooking safety factors by route type	Revenue Management Analyst	Amadeus Revenue Management (NRM)	Updated elasticity curves; historical no-show and cancellation rates by route category (short-haul leisure, long-haul corporate, thin route)	Revised overbooking multipliers per route cluster; updated denied boarding risk scores	Denied boarding rate ≤0.05 per 1,000 passengers; seat spoilage rate ≤2% of available seats	Y	Y
Phase 4 4.1	Execute shadow-mode simulation on hold-out period	RM Systems Analyst	Amadeus Revenue Management (NRM)	Updated NRM parameter set (demand model + elasticity curves + overbooking factors); hold-out historical period (most recent 90 days not used in training)	Shadow revenue report: simulated revenue and load factor for each O&D flight in hold-out period	Shadow simulation runtime ≤8 hours; simulated revenue variance vs. actuals ≤2%	N	N
4.2	Evaluate shadow revenue vs. actual revenue	Revenue Optimization Manager	Tableau	Shadow revenue report; actual revenue actuals from AWS Redshift for the same hold-out period	Calibration validation dashboard: revenue lift by market cluster, load factor delta, booking class mix comparison	Simulated revenue uplift ≥1.0% vs. actuals across all market clusters; no individual cluster showing revenue degradation >0.5%	N	N
4.3	Run live A/B test on selected pilot markets	Revenue Management Analyst	Amadeus Revenue Management (NRM)	Validated calibration parameter set; pilot market selection (10–15 O&D pairs representative of route clusters)	A/B test results: revenue per available seat kilometre (RASK) and load factor for treatment vs. control group over 4-week period	A/B revenue uplift ≥1.5% in pilot markets over 4-week test window at 95% statistical confidence	Y	Y
Phase 5 5.1	Stage calibrated parameters in UAT environment	RM Systems Engineer	Amadeus Revenue Management (NRM)	Approved calibration parameter set from A/B test; UAT deployment checklist	UAT environment loaded with new parameters; regression test results confirming no unintended impacts to adjacent processes (pricing, inventory, interline)	UAT regression test pass rate 100% on critical path; deployment staging completed within 1 business day	N	Y
5.2	Present calibration results for RM leadership sign-off	Head of Revenue Management	Microsoft Azure Synapse	Calibration validation dashboard; A/B test results; UAT regression summary; estimated annual revenue impact	Signed approval record or change request with required amendments; go/no-go decision for production deployment	Sign-off decision turnaround ≤2 business days; zero production deployments without documented approval	Y	N
5.3	Deploy calibrated parameters to production NRM	RM Systems Engineer	Amadeus Revenue Management (NRM)	Signed approval record; staged parameter set from UAT; deployment runbook	Production NRM updated with new demand model coefficients, elasticity curves, and overbooking factors; deployment confirmation log	Production deployment completed within 4-hour maintenance window; zero unplanned system downtime during deployment	N	Y
5.4	Monitor live KPIs post-deployment	Revenue Management Analyst	Tableau	Live booking data from Amadeus NRM; post-deployment KPI dashboard; baseline metrics from pre-calibration period	30-day post-deployment performance report; rollback recommendation if KPIs deteriorate beyond threshold	RASK improvement ≥1.0% vs. pre-calibration baseline within 30 days; no market showing load factor decline >2 percentage points	N	Y

📋

Process Attributes

Identification

Process IDNP-RM-08

L1 DomainNetwork Planning & Scheduling

L2 ProcessRevenue Management

L3 NameRM System Calibration & Tuning

L4 Steps16 across 5 phases

Decision Gates5 (all with iteration loops)

Exceptions6 documented

Swim Lanes (Roles)

Revenue Management Analyst

Senior Revenue Analyst

Revenue Optimization Analyst

Data Engineer

Data Quality Analyst

RM Systems Analyst

Revenue Optimization Manager

RM Systems Engineer

Head of Revenue Management

Systems & Tools

Amadeus Revenue Management (NRM)TableauAWS RedshiftAWS S3ATPCOMicrosoft Azure Synapse

Key Performance Indicators

Extract RM forecast accuracy metricsMean Absolute Percentage Error (MAPE) ≤15% across all O&D markets

Identify underperforming O&D markets≥90% of O&D markets within MAPE threshold within 30 days of calibration cycle

Assess booking curve deviations from expected pickupBooking curve R² ≥0.92 across flagged markets; decision to proceed with recalibration if ≥3 markets fail threshold

Extract historical booking and ticketing dataDataset completeness ≥98% of scheduled departures in scope; extract runtime <2 hours

Validate input data quality and completenessData quality score ≥95% (null rate <2%, duplicate bookings <0.5%, fare outliers <1%)

Load competitive fare and capacity benchmarksCompetitive data coverage ≥85% of target O&D markets; data freshness within 48 hours of filing date

Recalibrate demand forecast model coefficientsPost-recalibration MAPE improvement ≥10% vs. baseline on hold-out validation set

Update price elasticity and WTP model curvesBid price accuracy: simulated revenue within ±2% of optimal on hold-out markets; elasticity curve R² ≥0.88

Airline-Specific Risks & Pain Points

Amadeus NRM MAPE reporting requires custom extract scripts; no native scheduled accuracy dashboard, increasing analyst effort and risk of metric inconsistency

Cross-market comparison requires manual joins between Amadeus NRM outputs and AWS Redshift actuals; data latency of up to 24 hrs can mask intra-day booking pattern shifts

Structural demand shifts post-COVID (early leisure surge, late corporate recovery) have permanently altered booking curves; NRM baseline curves calibrated pre-2023 may no longer reflect current market behaviour

AWS Redshift query performance degrades on unpartitioned historical tables; full 24-month O&D extracts for 100+ markets can exceed 4-hour SLA without query optimisation

Amadeus Altéa PSS ticket voids and exchanges frequently generate duplicate or partial booking records that inflate cancellation rate inputs to the NRM demand model, distorting recalibration outputs

ATPCO fare data covers only filed public fares; negotiated corporate and NDC-only fares are not visible, creating a systematic blind spot in price elasticity model inputs for corporate-heavy routes

Inputs / Outputs

Primary InputScheduled monthly calibration cycle or ad-hoc trigger from revenue variance alert

Primary Output30-day post-deployment performance report; rollback recommendation if KPIs deteriorate beyond threshold

←PreviousNP-RM-07 · Ancillary Revenue Optimization NextNP-RM-09 · Post-Departure Revenue Reporting→