Deep diveTravel
Airline dynamic pricing under NDC: from filed fares to offers and orders
Airline dynamic pricing under NDC means the airline builds each offer itself, at the moment of the shopping request, instead of publishing a fixed ladder of fares for intermediaries to assemble. NDC is the messaging standard that carries those offers to sellers; continuous pricing and dynamic bundling are what the airline does inside its offer engine. This page explains the mechanism, where machine learning fits and the guardrails it needs.
On this page
- How airlines sold seats before offers: fares, booking classes and tickets
- Vocabulary of airline offer and order management
- Filed-fare distribution compared with airline-built dynamic offers
- One shopping request traced from seller to order
- Where machine learning sits inside an offer engine
- Guardrails a dynamic offer engine needs before launch
- A hypothetical flexible bundle for a business trip
- Connecting the offer engine to PSS, revenue management and partners
- Questions and answers
- Sources
How airlines sold seats before offers: fares, booking classes and tickets
In the traditional model, an airline files fares and their conditions through a fare-filing intermediary, with ATPCO as the main channel, and links each fare to a booking class. Its revenue management system then decides how many seats to release in each class on each flight. Global distribution systems and agency tools combine the published fare with seat availability to price an itinerary, so the price a traveler sees is assembled outside the airline from parts the airline set in advance.
Fulfilment followed the same split. The reservation lived in the passenger name record (PNR), the right to fly in an electronic ticket, and paid seats, bags or lounge access in separate electronic miscellaneous documents (EMDs). With a fixed ladder of fare levels, the airline could only open or close rungs, never quote a price between them or shape a bundle around the request in front of it.
Vocabulary of airline offer and order management
These terms are used loosely in vendor material. Pinning them down makes it easier to see which system owns which decision.
- NDC (New Distribution Capability)
- An IATA program and XML-based data exchange format built on offer and order management, which any airline, intermediary or IT provider can implement1.
- Offer
- A priced, time-limited proposal the airline builds in response to one shopping request. It can combine flights, fare conditions and ancillaries, and it expires if not accepted.
- Offer management system
- The airline-side engine that decides which products to include in an offer, at what price and in what order, using inventory, revenue management controls and merchandising rules.
- Continuous pricing
- Quoting any price point rather than only the fixed levels of a filed fare ladder, usually as an adjustment relative to a reference fare.
- Bid price
- The revenue management estimate of a seat's opportunity cost on a flight or network path. An offer that prices below it gives away expected future revenue.
- Dynamic bundle
- A package of flight and ancillaries assembled and priced for the request, as opposed to a branded fare family fixed in advance.
- Order (ONE Order)
- IATA's XML-based standard that combines booking and ticketing records into a single customer-focused order, gradually phasing out separate PNRs, e-tickets and EMDs2.
Filed-fare distribution compared with airline-built dynamic offers
| Aspect | Filed fares and booking classes | Dynamic offers over NDC |
|---|---|---|
| Who assembles the price | The GDS or agency system, from filed fares and availability | The airline's offer engine, per request |
| Price granularity | Fixed fare levels opened or closed by booking class | Any price point above the airline's floor |
| Ancillaries in indirect channels | Limited, often sold separately after booking | Bundled or offered alongside the flight at shopping time |
| Content consistency by channel | Largely uniform, because everyone prices the same filed data | Depends on the airline's own rules; divergence is possible and must be governed |
| Servicing after sale | Reissue or revalidate tickets and EMDs | Change the order; the airline reprices the difference |
| Where the airline loses control | Display, ranking and bundling by intermediaries | Aggregator and seller tools that may not render rich content well |
Most airlines run both models side by side for years; the right column describes the target state, not a switch-over date.
One shopping request traced from seller to order
The exchange below shows where pricing models are called. Message names vary by NDC schema version, so the labels describe purpose rather than exact message types.
- Seller tool
An agency desktop, corporate booking tool or the airline's own website.
- Aggregator
Optional intermediary that connects many sellers to many airline NDC endpoints.
- Airline offer engine
Builds candidate offers from inventory, bid prices and the product catalog.
- Pricing and ranking models
Estimate willingness to pay and the likely choice for this request context.
- Order management
Creates and services the order and passes data to accounting and delivery.
- Seller tool to Aggregatorshopping request
- Aggregator to Airline offer enginerequest + context
- Airline offer engine to Pricing and ranking modelscandidate offers
- Pricing and ranking models to Airline offer engineprices and ranks
- Airline offer engine to Seller tooloffers with expiry
- Seller tool to Airline offer engineconfirm price
- Seller tool to Order managementcreate order
- Order management to Seller toolorder confirmed
Where machine learning sits inside an offer engine
Demand forecasting does not go away. The offer engine still depends on revenue management forecasts by market, date and days before departure, and on the bid prices they produce. Dynamic offers sit on top of that control layer rather than replacing it.
Willingness-to-pay estimation adjusts the price of an offer relative to its reference fare, using the context of the request: advance purchase, trip length, whether a weekend is included, party size, point of sale and channel. These signals describe the trip, not the person, which keeps the model closer to classic fare-rule logic and easier to defend. The research behind such estimates draws on the same methods as willingness-to-pay research for new products.
Offer construction and ranking decide which bundles to include and in what order. A model can learn, for example, that requests with short advance purchase and no weekend stay often choose flexibility, and lead with a flexible bundle. Experimentation closes the loop: controlled tests with holdout groups, judged on revenue per shopping request and conversion together, because a price that converts less but earns more per sale may still lose money overall.
Guardrails a dynamic offer engine needs before launch
Displayed prices that break final-price rules
Early signalTaxes, carrier charges or pre-selected extras appear only late in the booking flow.
MitigationIn the EU, Article 23(1) of Regulation (EC) No 1008/2008 requires the final price, including unavoidable and foreseeable taxes and charges, to be shown at all times, with optional supplements accepted on an opt-in basis3. In the US, 14 CFR 399.84 requires advertised prices to be the entire price to be paid4. Build these rules into offer assembly, not just the website.
Personalization that becomes a disclosure obligation
Early signalModel features start to include identity, loyalty history or device signals.
MitigationDirective (EU) 2019/2161 added a duty to tell consumers when a price was personalized on the basis of automated decision-making to the EU Consumer Rights Directive5. How far it reaches passenger transport is a legal question, so record which features each model uses and take advice before adding personal signals.
Channel divergence that damages agency and corporate trust
Early signalTravel managers report different prices for the same flight in different tools.
MitigationDefine which differences are intended, such as negotiated fares, and test the rest automatically across channels.
Revenue managers who cannot explain or override the engine
Early signalOverride rates climb and analysts rebuild prices in spreadsheets.
MitigationShow the reference fare, bid price and model adjustment for every offer, and give analysts bounded overrides with an audit trail.
Models trained on demand that no longer exists
Early signalOffers on a route behave oddly after a schedule change, a competitor entry or a disruption period.
MitigationMonitor forecast error by market, cap adjustments where error rises and retrain with the abnormal period flagged rather than silently included.
A hypothetical flexible bundle for a business trip
Connecting the offer engine to PSS, revenue management and partners
The passenger service system remains the source of truth for seat inventory, so the offer engine must read availability and write holds without creating a second inventory. Revenue management supplies forecasts and bid prices, the merchandising catalog supplies ancillaries, and finance needs orders that revenue accounting can recognize.
Most airlines will keep traditional GDS distribution alongside NDC for a long time, and interline partners still exchange tickets in many cases. Run the same pricing logic behind both paths wherever possible, or the two channels drift apart. ColdAI's travel practice designs integrations with GDS, CRS and other distribution systems, and older PSS code can be wrapped and replaced in stages using approaches such as the strangler fig pattern.
Questions and answers
Is NDC the same thing as airline dynamic pricing?
No. NDC is a data exchange standard that lets an airline send its own offers to sellers and receive orders back. Dynamic pricing is a commercial capability inside the airline's offer engine. NDC makes dynamic offers easier to distribute through agencies, but an airline can connect over NDC while still pricing from a filed fare ladder, and it can price dynamically on its own website without NDC.
Does dynamic offer pricing mean each passenger gets a personal price?
Not necessarily. Most dynamic offer models adjust prices using the context of the trip, such as advance purchase, trip length and point of sale, rather than the identity of the traveler. Using personal data to set prices raises separate consumer-law and data-protection questions, including EU disclosure rules for prices personalized by automated decision-making, so it should be a deliberate, documented decision.
How do travel agents see airline NDC offers?
Through an agency desktop, corporate booking tool or GDS that has built an NDC connection, either directly to the airline or through an aggregator. What the agent sees depends on how well that tool renders rich content such as bundles, seat maps and images. Servicing support, such as changes and refunds on NDC orders, also varies by tool, so agencies often check it before moving volume.
Sources
- New Distribution Capability (NDC) — IATA · checked 10 October 2026
- ONE Order — IATA · checked 10 October 2026
- Regulation (EC) No 1008/2008 on common rules for the operation of air services in the Community (Recast) — EUR-Lex · checked 10 October 2026
- 14 CFR 399.84 Price advertising and opt-out provisions — Legal Information Institute, Cornell Law School · checked 10 October 2026
- Directive (EU) 2019/2161 as regards the better enforcement and modernisation of Union consumer protection rules — EUR-Lex · checked 10 October 2026