This is the second of three essays on the changing shape of airline commercial intelligence. The first argued that competitive fare benchmarking remains the daily discipline, but that continuous pricing and fragmenting products are dissolving the snapshot underneath it. This essay takes up the imminent wave: the unit of comparison moving from a filed fare to a constructed offer, and demand becoming readable from search and channel - before any price is shown.
The catalyst is not a new idea about retailing. It is a change in what indirect channels can carry. Under EDIFACT, the airline saw a filed fare, a class, and a coarse outcome. Direct channels already had finer grain: first-party search, a live offer, and a result. NDC and Offer and Order put that same grain into the indirect path. House splits on the order of 60 per cent direct and 40 per cent indirect are a reminder of why that matters. A large minority of the business is no longer visible only as a filed ladder. It can be visible as an offer assembled at shop time, with the search that preceded it.
Once that is true, two jobs open at once. The commercial team can compare the whole product, not the headline fare. It can also treat intent as a measured signal rather than as exhaust from the booking.
Competition is moving from a single fare to the full offer
The older comparison was simple because the object was simple. One published price, one booking class, the same product in every channel, a ladder the market could see. Benchmarking meant pulling the fare and ranking it.
The object is no longer simple. Price arrives with brand, bundle, and ancillary, assembled at the moment of shop. Channels receive different content and different prices. A competitor may never file the offer that a traveller actually sees. Ranking the lowest available fare across that landscape is a ranking of unlike products unless the tiers have been normalised first.
That is a retailing problem wearing a data problem’s clothes. Airlines have spent a decade telling the market they are becoming retailers. Retail comparison requires a common shelf. “Basic” on one carrier is not automatically “Light” on another. Seat selection, baggage, change privileges, and lounge access move in and out of the fare by market, by channel, and by passenger mix. Until those attributes sit in a shared map, a pricing meeting that debates a two-dollar gap is debating a number that may be attached to two different products.
Channel-specific construction makes the map more necessary, not less. Direct, TMC, OTA, and retail agency are not four windows onto one offer. They are four construction environments. A filed-fare file cannot say whether an indirect weakness is a price problem or a merchandising problem. An offer-level file can, but only after brand names and bundle elements have been forced into comparable tiers. The unglamorous work described in the first essay (normalise, then rank) is the price of using the finer grain NDC makes available.
Search is demand, not exhaust
Price, even a well-normalised offer price, is a lagging description of who bought. Search is a leading description of who is still looking. City pair, dates, and passenger mix are visible in the request before any fare is returned. That is real-world demand, not demand inferred from the ticket that happened to clear.
The idea is not new. Analysts have treated search as a leading indicator for years. What is new is that the signal is becoming usable at commercial scale, including in indirect, once Offer and Order carries first-party grain beyond the airline’s own site. A PROS study of 27 airline clients found a Pearson correlation of 0.77 between flight search initiation and purchases. A ten per cent rise in search initiation tracked to an approximately 7.3 per cent rise in purchases, significant at p < 0.01 (PROS, 2025). That is not a metaphor about “intent.” It is a measured relationship between looking and buying.
Two cautions belong next to that result. Correlation is not a complete demand model, and search on a single carrier’s site is not the market. The study is still sufficient for the claim this essay needs. Search volume, by market and by passenger mix, is a finer reading of who remains in the funnel than the fare that closed yesterday. Commercial teams that treat look traffic only as a cost, the look-to-book problem in the first essay, are throwing away the information the cost produces.
Channels are different markets
The same grain has a second cut: channel. Direct, TMC, OTA, and brick-and-mortar retail do not sample the same traveller. They do not respond to the same events. A single market number averaged across them hides the movement that revenue management, network, and sales actually have to act on.
The pattern is familiar even before any one carrier’s house file is opened. Direct mixes brand-loyal demand with metasearch spill and tends to show a slight yield premium. TMC demand is managed corporate traffic: policy, duty of care, shorter booking windows, international and long-haul weight, a clearer premium-cabin skew, and a material yield premium. OTA demand is price-led leisure, more domestic, more economy, longer advance purchase, and a yield discount. Retail agencies sit between those poles; advice-led leisure, more international than the OTA, mixed cabin, a smaller discount. House metrics used in the underlying briefing put illustrative yield indices at 1.05 direct, 1.20 TMC, 0.80 OTA, and 0.90 retail against an airline average of 1.00. The levels will differ by carrier. The structure does not.
Advance-purchase and entity mix make the same point. A fare action that is rational against OTA leisure six to twelve weeks out can be the wrong action against TMC demand one to three weeks out on the same city pair. Network decisions that treat “the market” as one series will misread a channel mix shift as a demand shift. Loyalty and marketing, looking only at cleared tickets, will miss where the looking has already moved.
Each channel is therefore a demand sample with utility beyond sales reporting. Revenue management can see whether a weakness is concentrated in price-led leisure or in managed corporate. Network can see whether a route’s apparent softness is an OTA phenomenon. Distribution can see whether an indirect construct is underselling a product that direct is clearing. Finance can see yield as a portfolio of populations rather than as one average. That is the operational meaning of “one grain, two jobs.” The same offer-and-search file that supports competitive comparison also supports a demand reading the older fare file could not produce.
One picture, used everywhere
The payoff is organisational rather than technical. Pricing, revenue management, network, and distribution have always needed the same competitive picture. They have not always been looking at the same object. Filed fares gave them a shared number and a shared fiction: that the number was the product. Offer-level data, once normalised, gives them a shared object that matches what the traveller was shown. Search-level data gives them a shared leading indicator. Channel cuts give them the populations inside the average.
None of that requires a new commercial question. It requires admitting that the question (where do we sit, and who is still looking) cannot be answered by a headline fare pulled on a cycle. NDC and Offer and Order do not automatically produce the answer. They produce the grain. The work is to keep the grain comparable across carriers and usable across functions.
The third essay turns to the wave after this one. Travellers are already using AI to plan and book. The rails for that behaviour are being built by platforms that are not travel consortia. New query types will not look like a GDS fill. The industry will have to decide whether the storehouse for those queries is a neutral layer or a new choke point. The offer-and-intent file described here is what that next layer will either extend or bypass.
