Southwest Folds: NDC Is Table Stakes Now

Eric Marketts

On May twenty-eighth, 2025, Southwest Airlines started charging for a checked bag. On January twenty-seventh of this year, it ended fifty-four years of open seating. And on August third, it announced NDC.

Steph Nell

Read those three dates in order and the story tells itself.

Eric Marketts

Say it anyway, because most of the coverage this week did not.

Steph Nell

Southwest did not adopt NDC because the standard finally won the argument. It adopted NDC because it finally built something that needs selling.

Eric Marketts

Welcome back to The V1 Airline Retailing Report. I'm Eric Marketts.

Steph Nell

And I'm Steph Nell. Last week we argued that the pipe stopped being scarce, and what is scarce now is the offer and the shelf it lands on. This week the last major U-S holdout in this industry stood up and proved the first half of that, and then walked straight past the second half without appearing to notice it was there.

Eric Marketts

For anyone new, here is what we do. Every Monday we take the developments that actually move airline and travel distribution and run each one through the bull case, the bear case, and the critical take. Not the press release. The strategic read. Sources are in the episode details.

Steph Nell

Three narratives this week, and as usual they are not a list. They are one argument.

Eric Marketts

Narrative one. Southwest Airlines commits to NDC, picks Amadeus Altea NDC, and builds its own direct connect at the same time. The last big holdout folds, and the reason it folded is not the reason you have been told.

Steph Nell

Narrative two. The Model Context Protocol has a production-scale foundation under it now. Travelport put its TripServices platform into more than four hundred European agencies through Travelsoft on July first. That is the deterministic layer an agent would call. Whether the agents are calling it yet is a different question, and nobody has published that number. Meanwhile the framing in the market has become MCP versus NDC, which is the wrong fight, described wrongly, by people who need you to pick a side.

Eric Marketts

Narrative three. IATA is pushing hard on one hundred percent Offers and Orders. A consortium readiness programme now in its third edition, seven full-scope airline contracts, forty-eight pilots running, and an aspirational 2030 date on the wall. The standards body is trying to convert a decade of framework into shipped systems.

Steph Nell

The through-line. Every one of these stories is about sequencing. Southwest shows you that distribution follows product. MCP shows you that the interface layer is being built by whoever moves first, not by whoever standardized first. And IATA shows you an industry finishing the back office of retailing while the front end is redefined by parties it does not control. Three stories about the same mistake: confusing the thing you finish with the thing you win.

Eric Marketts

Finish the standard. Just stop calling it a strategy. Let's get into it.

Eric Marketts

Before the news, help me set the table for anyone who has not spent twenty years inside airline distribution. What was actually different about Southwest?

Steph Nell

For most of its history Southwest was the cleanest argument in the industry against intermediated distribution. One cabin. No assigned seats. No first bag fee, no second bag fee. Fares that were simple enough to publish on a wall. If your product is one undifferentiated seat at a transparent price, you do not need a sophisticated retailing channel, because there is nothing to retail. You need a booking channel. So Southwest built one, called it Southwest dot com, ran about eighty percent of its passenger revenue through it for years, and told the travel agency world to come to the website. That number is seventy-eight percent now and falling, because Southwest added Expedia and Priceline last year. Even the direct-channel purist is diluting the direct channel on purpose.

Eric Marketts

And it was not that Southwest could not get into the GDS. It chose not to.

Steph Nell

Mostly, yes, and it was rational. Every dollar of distribution cost you avoid is a dollar of fare advantage you keep. That was the whole model. It softened in 2020, when Southwest put content into Travelport, then Amadeus, and signed Sabre at the end of that year. It has run SWABIZ for corporate bookers for years. But that content went out on the old plumbing, and it went out flat. Fares and schedules. A ticket, not an offer.

Eric Marketts

So define NDC for me one more time, in the way that matters for this story.

Steph Nell

NDC is the messaging standard that lets an airline construct its own offer and push it out through a third party, rather than handing over a fare filing and hoping the intermediary assembles something coherent. It is how an airline says: for this traveler, on this route, today, here is the seat, the bag, the boarding position, the bundle, and the price of the whole thing together. You only need that if you have things to combine.

Eric Marketts

Right. And that is the setup. Here is what happened this week.

Eric Marketts

On Monday, August third, Southwest announced two things in one release. First, a corporate product called Business Priority, launching early 2027, giving eligible business travelers preferred re-accommodation, standby, and boarding benefits, sold as an optional add-on to corporate travel agreements. Second, and this is the one the industry noticed, NDC connectivity. Southwest is evolving its own in-house API for direct connections, expected to be available to partners by the end of 2026. And it has selected Amadeus Altea NDC to deliver what the release calls personalized and enriched offers through third-party distributors, including the Amadeus travel platform. That one lands later in 2027. Aileen Furlong, VP of Sales at Southwest Business, framed it as making it easier for corporate buyers to access the enhancements through expanded distribution options.

Steph Nell

Two channels, two timelines, one carrier. The direct API inside five months. Amadeus a full year behind it. That is not a rollout. That is a portfolio.

Eric Marketts

Give me the bull case, because I think there is a genuine one here and it is not just "the standard won."

Steph Nell

There is a real one. Corporate demand forced this, and corporate demand is the right forcing function. Travel managers have wanted Southwest content inside their booking tools with the same fidelity as every other carrier for a decade. Now they get fares, seat products, and ancillaries in the workflow their travelers already use, instead of a bolt-on portal that nobody outside of the travel department remembers exists. Amadeus gets a marquee reference for Altea NDC, which matters because Altea NDC has needed a headline name in the U-S. And the industry gets to close a long-running argument. When the carrier whose entire brand was built on rejecting intermediated distribution builds an NDC channel, the debate about whether NDC is real is over. It is over on the merits, not on the marketing.

Eric Marketts

And where does it break down?

Steph Nell

Start with the timeline, because airline distribution roadmaps slip as a rule. Direct API by end of 2026 is five months away and Southwest is simultaneously absorbing assigned seating, extra legroom, bag fees, a new fare bundle structure, and an international partner program. Altea NDC "later in 2027" is a fifteen-month horizon in an industry where fifteen-month horizons routinely become thirty. Second, running a proprietary direct connect and a GDS-mediated NDC channel at the same time doubles your integration surface, doubles your certification burden with every seller, and puts you in the channel-conflict business permanently. You now have to decide, forever, what your own API gets that Amadeus does not. Third, and this is the one nobody says out loud: Southwest's cost advantage was partly a distribution cost advantage. NDC through a GDS reintroduces transaction economics the carrier spent thirty years engineering around.

Eric Marketts

So give me the critical take, because I do not think the take is "the holdout capitulated."

Steph Nell

The take is that everyone has the causation backwards. Read the dates. May 2025, Southwest starts charging for the first checked bag. January twenty-seventh, 2026, open seating ends after fifty-four years and assigned seating and extra legroom go lye'v. Then fare bundles. Then Business Priority. Then, on August third, NDC. Southwest did not spend a decade being persuaded that NDC was a good standard and finally come around. Southwest spent a decade with a product that did not require it, changed the product, and then discovered it had no channel that could sell what it had built. You cannot express a seat assignment, an extra-legroom upsell, a bundle, and a priority boarding entitlement through a flat fare feed. You need an offer. So it went and got one.

Eric Marketts

Which changes what we should be measuring.

Steph Nell

Completely. The whole NDC adoption scoreboard is built on the assumption that adoption is a function of standards maturity, vendor readiness, and industry evangelism. It is not. It is a function of product complexity. Airlines adopt NDC at exactly the moment their product gets complicated enough that they lose money without it, and not one quarter before. Which means if you want to know which carrier commits next, stop reading the IATA adoption charts and go look at who just unbundled, who just added a premium cabin, who just launched a subscription, who just started selling seat products. That is your pipeline. The standard is downstream of the merchandising decision, and it always was.

Eric Marketts

And the corollary is unkind to a lot of roadmap decks.

Steph Nell

It is. If you are an airline with a simple product and you are running a three-year NDC program because the industry says you should, you have bought a retailing channel for a thing you are not retailing. The order is: change the product, then change the pipe. Not the reverse. Southwest just ran the correct sequence in plain sight, and the industry read it as a conversion story.

Eric Marketts

Hold onto that, because story two is about a different layer entirely, and the same sequencing mistake shows up there wearing better clothes.

Steph Nell

And notice what is not in the Southwest release. Two channels, two timelines, a corporate tier, a full distribution architecture published through 2027. Not one word about AI agents. Keep that in your pocket.

Eric Marketts

Walk me through what the Model Context Protocol actually is, for someone who has heard the acronym in four vendor decks and still could not define it.

Steph Nell

It is a convention for how an AI model calls an outside system. That is it. A large language model is very good at producing plausible text and structurally incapable of knowing what a seat costs right now. Andrew Jordan, Travelport's chief product and technology officer, has been making the structural version of this argument all year. A large language model predicts the next probable token. It does not query lye'v inventory. When a model says a seat is available, that is an inference, not a lookup. MCP is the wiring that lets the model stop guessing and go ask. The model proposes, the actual system disposes.

Eric Marketts

So if NDC is the language airlines and sellers use to talk about an offer, MCP is the language a machine uses to reach in and do something with it.

Steph Nell

Close enough to be useful, with one correction that matters. NDC is a domain standard. It knows what a fare is, what an ancillary is, what an order is. MCP knows none of that and does not want to. It is plumbing. It is content-agnostic. Anyone can wrap anything in it in an afternoon, which is its entire strength and the whole reason it moved this fast.

Eric Marketts

And that is the setup. Here is the news.

Eric Marketts

Travelport formally launched TripServices on June eleventh, and says it is backing its next growth phase with fifty million dollars. That is Travelport's own number and it is company-level, not a line item on this product, so take it as a statement of intent. Cognizant and Anthropic came in on May twenty-seventh, with Cognizant supplying the engineering muscle to move decades of fare logic into the new architecture. Then on July first, the Travelsoft integration went lye'v: more than four hundred travel agencies across Europe reached through Travelsoft's Orchestra platform and Travel Compositor, with rollout continuing to airQuest, Atcore, Tigerbay, Traffics, and Travel Connection Technology. And the scope is not a search toy. It covers exchanges, involuntary changes, ancillaries, ticketing, upsell, and multi-city. One precision point, because it matters. Neither Travelport nor Travelsoft used the word MCP in that announcement. The deterministic layer is there. The MCP framing on top of it is analysis, including ours.

Steph Nell

So one GDS has the execution layer live across four hundred agencies. Not the agents. The layer the agents would call. And for a rough read on how far this goes, Gimmonix, citing a study from Aven Hospitality and h2c, puts the share of hotel organizations that can actually complete an agent booking in real time at about eleven percent. That is a vendor's number, and it is about hotels, so hold it loosely. Nobody has published the equivalent number for airlines, which is its own answer.

Eric Marketts

Bull case first. Why is this the right architecture?

Steph Nell

Because it solves a problem NDC never tried to solve, and it solves it without another ten years of standards work. NDC answered one question. What is the offer. It never said how a machine buyer proves who it is, holds a seat, pays for it, and finds out whether the ticket actually issued. That was never in scope, and it should not have been. An MCP layer sitting on top of the NDC and EDIFACT plumbing you already have lets the agent economy plug into distribution as it is, today, without asking three hundred carriers to agree on anything. And here is the part that matters. The model guesses. The booking system knows. MCP keeps those two jobs apart, so the model never gets to invent a seat that is not there. It has to go and ask. That is a real piece of engineering and it deserves credit.

Eric Marketts

And the bear case.

Steph Nell

MCP is an integration convention. It is not a settlement framework, a ticketing framework, or a liability framework. Wrapping thirty years of accumulated distribution logic in a clean interface does not fix the inventory, refund, or interline behavior underneath. It hides it from the agent until something breaks, and then the agent has no vocabulary for what broke. And there are early warning signs, though I want to be honest about how thin the record is. A SmartCustomer analysis of consumer reviews in June found a pattern of AI booking platforms marking pending bookings as confirmed and charging cards without authorization. That is review mining, not a regulator, and nobody has named a platform. But it is the shape of the failure, and it is showing up before the volume does. Then there is the demand side, which the vendor decks skip. Expedia surveyed more than five thousand seven hundred adults in March. Fifty-three percent are comfortable letting AI suggest options. Eight percent are comfortable booking through an AI platform. Eight. And note who ran that survey, because Expedia has every reason to like that answer. Even discounting for it, the constraint is not architecture. It is that almost nobody will let a machine spend their money yet.

Eric Marketts

So what is the critical take?

Steph Nell

The framing is a category error, and it is a load-bearing one. "MCP versus NDC" is like asking whether English beat the telephone. NDC is what you say. MCP is how you reach the other party. They do not compete, they compose, and the reason the versus framing keeps getting published anyway is that it is commercially useful to whoever is standing next to it.

Eric Marketts

Useful how?

Steph Nell

If you can convince an airline that MCP and NDC are rivals, you can convince it that the decade and the budget it spent on NDC were a bet on the losing horse, and that it now needs to buy the agent layer from you rather than build it. That is the sale. And the airline that believes it makes a very specific error: it concludes it does not need to own the agent interface, because the agent interface is somebody else's protocol. Wrong conclusion. The protocol is nobody's. It is open, it is trivial to implement, and it will be commodity within eighteen months. What is not commodity is what sits behind your endpoint. Who is allowed to call it, how often, at what price, with what data, under what terms.

Eric Marketts

Which is the thing we have now asked for twice on this show.

Steph Nell

Twice, and I will say it a third time, because it is the whole game. In episode eight we said look-to-book economics were breaking under AI agent traffic, and we asked for the first carrier to publish an explicit AI traffic policy. In episode fifteen we gave partial credit to Air France-KLM for filtering what an agent gets back. I want to correct that on air, because we went back and checked it this week. That filtering is Amadeus's Advanced Airline Profile, it was announced in December 2025, and it targets unproductive NDC shopping traffic generally, not agents. So it was not movement, and it was not new. Partial credit withdrawn. Which leaves the ask exactly where episode fifteen put it, with nobody having met any part of it: authentication, rate limits, and a price for the query. Somebody will point at Delta's AI terms of use page. I went through it. It governs Delta's own chatbots and it bans scrapers that overload the servers. That is a consumer notice, not an access regime. And now look at Southwest. A carrier publishing a complete distribution architecture out to 2027, direct API and GDS NDC both, and there is not one word in it about who or what is allowed to call that API. The most technically ambitious distribution announcement of the year is silent on the buyer that is actually growing.

Eric Marketts

So the two stories are one story.

Steph Nell

They are. Southwest is building the offer layer that last week's episode said was the scarce thing. Correct move, and I want to be fair about that. But it is building it for the buyer it has, not the buyer it is getting. Meanwhile Travelport is building the shelf that machine buyer will actually call, in production, across four hundred agencies, today. Both of those are true at once, and only one of them is being measured by anybody.

Eric Marketts

Third story is the standards body. For anyone who tunes out when IATA comes up, tell them why the plumbing matters.

Steph Nell

Because the plumbing is why your change fee experience is terrible. A trip today does not Liv in one clean place. There is a PNR, which is a reservation record. There is an e-ticket, which is a separate financial document. There are EMDs for the extras, which are separate again. Three artifacts, three systems, three reconciliation paths, all invented between the 1960s and the early 2000s. Offers and Orders replaces that with one Order, which holds everything the customer bought and everything they are owed, in one object you can change without a fare rule argument.

Eric Marketts

And the industry has been at this for how long?

Steph Nell

NDC launched in 2012. ONE Order followed. We are fourteen years in. And here is the number most people still get wrong, because they are quoting a figure from 2022. ARC settled twenty-one point two percent of its transactions through NDC in December 2025, up from twenty point three a year earlier. More than one in five agency bookings. So NDC is not stalled. But look at who is doing it. OTAs are seventy-seven percent of that volume. Corporate agencies are seven. The channel that needs the offer most is the one that has it least. That is the number behind everything I am about to say.

Eric Marketts

Here is what moved.

Eric Marketts

The Airline Retailing Consortium's readiness work is now in its third edition, and it is aimed squarely at moving the industry off the legacy artifacts, e-ticket, PNR, and EMD. The first paper surveyed fifteen I-T providers on whether they could actually deliver Offers and Orders. The 2025 update surveyed twenty, and every one of them backs modular, interoperable architecture. As of last October there were seven full-scope airline contracts and forty-eight pilots running, and the first legacy retirements are scheduled from 2027. The Consortium started in late 2022 with ten carriers. American, Air France-KLM, British Airways and Iberia, Emirates, Finnair, Lufthansa Group, Oman Air, Singapore, and Xiamen. It has since added LATAM, Qatar, Turkish, Air India, Air Canada, and United. And look at what it is actually working: modular architecture that decouples the order layer from the monolithic PSS, interline retailing through the SRSIA framework, order-centric disruption management, and Settlement with Orders. On the value side, IATA's own vision paper leans on a McKinsey estimate of up to seven dollars of additional value per passenger by the end of this decade. Industry average. The date on the wall is 2030, and IATA's own word for that date is aspirational.

Steph Nell

Twenty providers surveyed, seven full-scope contracts, forty-eight pilots. That is a real signal and I do not want to be cynical about it. That evidence is the part that has been missing for a decade.

Eric Marketts

Bull case.

Steph Nell

The industry finally needs the tangibles, and this is what tangibles look like. Offers and Orders only pays off when the PNR and the e-ticket actually retire and one Order carries the customer end to end, through servicing, changes, refunds, and ancillaries. Until then you are running two worlds and paying for both. Moving from principle to migration playbook, with vendor readiness assessed and named, is exactly what de-risks this for the mid-tier carrier that cannot afford to be first. Settlement with Orders is the sleeper item on that list, by the way. Order-based financial clearing shortens the cash cycle, and a CFO will fund a cash cycle improvement long after they have stopped funding a retailing vision. And seven dollars a passenger is not a rounding error at scale.

Eric Marketts

Bear case.

Steph Nell

One hundred percent is a decade-late aspiration meeting a moving target. Airlines are being asked to complete a PSS-deep transformation whose customer-facing surface may be commoditized before the migration finishes. Every quarter spent retiring the PNR is a quarter not spent on the interface the buyer is actually arriving through. And read that workstream list again. Modular architecture, interline, disruption, settlement. Every one of those is internal. Necessary, all of them, and not one of them says a word about how an autonomous agent authenticates against an Order, what it is entitled to see, or what it pays to ask. The roadmap has a 2030 date and a buyer profile from 2012.

Eric Marketts

Critical take.

Steph Nell

IATA is standardizing the back office of retailing at the exact moment the front end is being redefined by parties outside its membership. And I want to be careful here, because the lazy version of this take is that Offers and Orders is a waste of time, and that is wrong. It is necessary. You cannot retail dynamically on a record structure designed for a paper ticket, and every carrier that skips it will hit the wall in five years. But the standard was designed for a world where the seller is a human-operated channel. If the buyer becomes an agent calling an endpoint, the Order still matters enormously as the airline's internal source of truth, and it matters much less as a distribution differentiator. Those two things got welded together in every business case written since 2018, and this week is pulling them apart.

Eric Marketts

So what does an airline do with that on Monday morning?

Steph Nell

Finish Offers and Orders. Fund it as an operational upgrade, defend it on cost, cash cycle, and servicing, and hold it to those numbers. Stop selling it internally as a customer acquisition strategy, because it is not one and it never was, and the day your board asks why the revenue lift did not show up is the day the program gets cut at seventy percent complete. That is the actual risk here. Not that Offers and Orders is wrong. That it is oversold, misses a target it was never built to hit, and dies two years from the finish line with the PNR still in production.

Eric Marketts

Which is the worst possible outcome, because you have paid for both worlds and finished neither.

Steph Nell

That is the trap, and it is not hypothetical. Ask anyone who lived through a halted PSS migration.

Eric Marketts

Here is the spine. All three stories are about sequencing, and all three show the industry getting the order of operations wrong in the same direction. Southwest got it right and nobody noticed: change the product, then change the channel. MCP is being sold as a rival to NDC when it is a layer on top, and the versus framing exists to sell you something. And IATA is finishing the back office on a 2030 clock while the front end gets built this year by people who are not in the room.

Steph Nell

If you are new to this industry and wondering why any of it matters, here is the short version. For thirty years the fight in airline distribution was about who controls the pipe between the airline and the traveler. That fight is over and the pipe lost. What matters now is who the airline is actually selling to, and for the first time in the history of this business, the answer might not be a person. Everything in this episode is the industry arriving at that realization at three different speeds.

Eric Marketts

And for the veterans, the sharper version.

Steph Nell

Southwest published a full distribution architecture out to 2027 this week, and it contains no position on machine buyers. That is not an oversight by Southwest. That is the industry standard, and it is the third episode in a row where we have gone looking for a carrier that has published agent terms and come back with nothing. Authentication. Rate limits. A price for the query. Fourteen years to standardize the offer, and we cannot find a carrier that has published any of the three as a distribution policy. So here is the metric I would actually hold in 2026. Not NDC penetration. Not Offers and Orders percentage complete. Ask instead whether your distribution roadmap names the buyer you will have in 2028, and if the answer is a TMC and an OTA and nothing else, you are building a channel for a customer who is being replaced.

Eric Marketts

If you are inside one of these programs right now, send this to whoever is writing the business case. Not to agree with us. To argue with the sequencing, out loud, before the money is committed. These are decisions being made this quarter, not next year, and the ones that go wrong go wrong quietly and expensively.

Steph Nell

And if you are the carrier that has published agent terms, or is about to, we want to hear from you. Seriously. We have asked three times now. Somebody is working on this and we would rather cover it than keep counting the absence.

Eric Marketts

I'm Eric Marketts.

Steph Nell

I'm Steph Nell. Thanks for listening.

Eric Marketts

Finish the standard. Just stop calling it a strategy. Stay sharp out there.