Airline revenue management that says why a flight is off its curve, and waits for your yes
One more point of revenue per flight, and no seat change posted without an analyst's yes.
See one case, screen by screen ↓Airline revenue management combines airline demand forecasting with seat inventory control: setting fare class availability, bid prices and overbooking per flight so each seat sells at the best price it can. Airline Revenue Management adds seven agents to that work: they watch every flight-date against its booking curve, explain each deviation with its sources, propose fare class and overbooking changes, re-check each proposal and review every decision after departure. The revenue management analyst decides, and nothing posts to inventory without her yes.
How to choose AI for airlines: a buyer's guide to agentic AI solutions →
Updated
An airline revenue management analyst at a network carrier owns thousands of future flight-dates. The RM system handles demand forecasting and seat inventory control — bid price control opens or closes fare classes, and it flags the flights that drift off their booking curve — but it never says why. A rival sells out, school holidays start on a Saturday, network planning swaps the aircraft: the analyst finds out by opening fare shops, the reservations system and last year's flight, ten to twenty minutes per flight-date.
Then the work has no memory. Overrides of the optimizer expire silently and are never checked against what the flight actually earned, so nobody knows which judgment calls paid. And overbooking set from history alone bumps holiday travelers who nearly all show up — on a flight out of the US, denied boarding compensation runs up to $2,150 a passenger.
Estimated split for one analyst's portfolio on a busy day before the nightly run, by hand and with the solution.
Seven agents watch every flight, explain each deviation, propose and check the change; analysts decide.
Thursday, October 15. Megan Alvarez, revenue management analyst for Corvane Airways' ORD Atlantic portfolio, has until 18:00 to decide what goes into tonight's 22:00 optimization run. Here is how she works with the agents, in the working solution.
Her home screen opens on what needs her: 2,184 flight-dates watched to Sep 2027, 37 off their curve with a cause found for 36, and 9 proposals for her worth an estimated $81,400. The Deviation explainer has already written to her: three flights on the holiday peak need her before the 22:00 run, starting with CV 412 to London on Dec 22.
“Morning, Megan. I checked 2,184 flight-dates overnight; 37 are off their curve and I found a cause for 36.”
Holiday peak, Dec 19–24, out of ORD to London, Frankfurt, Paris, Amsterdam and Dublin: 30 slim strips, bookings against the forecast curve and its band. CV 412 on Dec 22 is 18% above its curve and CV 88 to Frankfurt on Dec 23 is 14% below, both outside the ±10% band; CV 64 to Paris on Dec 19 is near full. Beside the board the agents want to change those 3 flights, and nothing posts without her yes.
“18% above its curve. 188 booked against 159 expected at 68 days.” The flight ran below its curve while Northmere discounted, then took 61 bookings in 48 hours, against 4 typical, once Northmere sold out its evening flight. UK school holidays start that Saturday. Megan clicks the second citation and reads the fare shop itself: Northmere economy “not available in any booking class (sold out)”.
The explainer never states a cause it cannot cite. When it finds none, it says so and asks the analyst.
The flight page shows the curve with the 48-hour surge, demand re-forecast from 274 to 301 for 268 seats, and the economy ladder class by class. The Bid price agent proposes closing K ($605) and Q ($535) and raising the bid price from $480 to $565: an estimated +$41,200, spill risk 6. The Checker re-computed it on its own — bid $563, +$40,900 — and agrees within $2.
“Decide by today 18:00, 4 hours before the 22:00 optimization run. Nothing posts without your yes.”
She types the question under the note. The agent answers with sources: at $565 the bid turns away 14 connecting requests worth less than that on this leg, mostly DEN–ORD–LHR at $512, because local H and M demand fills those seats at about $905. At +$60 ($540) it turns away 9, keeps an estimated +$37,600 and lowers spill risk from 6 to 4.
The O&D displacement table lists every connecting itinerary on the leg, its requests forecast and its value: DEN–ORD–LHR, STL–ORD–LHR, MSP–ORD–LHR and more are accepted at $480 and turned away at $565, while CVG and OMA connections still clear. Below it, “How the agents got here” replays each step with its time, from the 06:02 flag to the 06:14 explanation.
“4 sources cited; no claim without a source.”
One click sets +$60; K and Q stay closed. The approval asks for a reason code — she picks CP, Competitor action — and for her name: the decision is hers, logged and checked after departure. Estimated effect +$37,600, spill risk 4, posted as soon as she signs.
“I, Megan Alvarez, make this decision under my name. It is logged and checked after departure.”
The Checker re-checks her edit on its own — +$37,300 against her +$37,600 — and agrees. The change posts to inventory and the inventory system confirms: K and Q closed, bid $540. The proposal reads “Edited · posted”, and every step, from her question to the posting, is on the audit trail.
CV 64 to Paris on Saturday Dec 19 has 204 of 211 seats booked with 5% overbooking. On holiday Saturdays 97.6% of ORD–CDG passengers show up, so 5% would bump an expected 3.8 people, at up to $2,150 each. The agent proposes 3%: an estimated +$4,300 net, with 0.7 expected bumped. The Checker agrees.
The Overbooking agent never goes above the cabin ceiling in Settings, and only proposes.
After departure the Outcome reviewer replays each flight with the old controls on the same demand. Last 30 days: 41 decisions checked, 33 helped, 6 even, 2 cost, an estimated +$612,000 net. It also spotted a pattern — after a rival sellout, Megan's trimmed bid raises earned more than the proposal 7 times in 9 — and suggests a rule for Daniel Brooks, the director, to accept or decline.
The Outcome reviewer proposes rule changes; it never changes a rule.
The same holiday peak, seen by four of the people who carry it — what their morning looked like, and what it looks like now.
Seven agents watch every flight, explain each deviation, propose and check the change; analysts decide.
Compares every flight-date with its booking curve, every hour, and flags those beyond the band set in Settings (±10%).
Reads fare shops, rival sellouts and schedule filings, events and aircraft changes, and links each signal to the flight-dates it touches.
Says why a flight is off its curve in one plain paragraph — the surge, the sellout, the holiday, the down-gauge — citing the bookings, shop, event or schedule behind each clause.
Re-forecasts unconstrained demand, computes O&D displacement and proposes the bid price and fare class availability — which classes to open or close — with the estimated effect and spill risk.
Forecasts show-ups and cancellations and proposes how many extra bookings to accept, with expected denied boardings and their compensation cost.
Re-computes every proposal from the raw bookings without seeing the proposer's number, and compares the two.
After each flight departs, replays it with the old controls on the same demand, says whether the decision paid, and suggests rule changes by reason code.
Ask why, edit, approve with a reason code; the director co-signs above the limits. The agents propose; a named person decides.
Ask about any flight, proposal or signal. Every answer cites its source and opens the screen it talks about.
Why is CV 412 on Dec 22 above its curve?
It took 61 bookings in 48 hours, against 4 typical, after Northmere sold out its evening flight to London, and school holidays in England start that Saturday. Demand is now 301 for 268 seats. The Bid price agent proposes closing K and Q and raising the bid $85.
Which proposals must I decide before tonight?
Four, by 18:00: P-1041 London Dec 22 (+$41,200), P-1042 Frankfurt Dec 23 (+$12,900), P-1043 Paris Dec 19 (overbooking 5% → 3%) and P-1047 London Nov 25, where the Checker differs.
What did my edits earn after departure?
In the last 30 days the Outcome reviewer checked 41 decisions: 33 helped, 6 even, 2 cost, an estimated +$612,000 net. Your trimmed bid raises after rival sellouts earned more than the proposals 7 times in 9 — for example CV 412 on Oct 11: +$9,400.
Where did a rival sell out or cut fares?
Northmere sold out economy on ORD–LHR Dec 22 and raised Dec 23 by $64. Aurelle Air sells ORD–FRA Dec 23 at $417, $118 under our lowest open fare.
13 screens from the working solution, on its sample data. Pick one to see it large.
What the agents may do alone, what needs the analyst's yes and what needs the director's co-sign are settings. Only the Director of Revenue Management changes them, and every change is logged.
| Setting | Default | Choose from |
|---|---|---|
| Analyst's bid price limit | ±$150 | $50 to $300, in $10 steps |
| Overbooking ceiling for the analyst | 5% | 1% to 10%, one point at a time |
| Close classes up to (days out) | 90 days | 30 to 180 days, in 10-day steps |
| Bulk approval up to | ±$25 | $5 to $50, in $5 steps |
| Flag a flight off its curve by more than | ±10% | 5% to 25% |
| Checker must agree within | $5 | $1 to $25 |
| Decide by, hours before the 22:00 run | 4 h | 1 to 8 hours |
| Block approval when the Checker differs | On | On or Off; when on, the analyst must read both workings and pick one first |
Every figure is labelled: a target the solution is built to, an estimate, a typical published result, or a proven one.
“demo” = seen in the working solution, on its sample data · “target” = the design goal, measured in the pilot · “estimated” = our estimate · “typical” = published figures (a revenue management vendor's published study, as reported by AltexSoft, linking a 10% gain in forecast accuracy to about 1% more revenue; 14 CFR 250.5 on denied boarding compensation up to $2,150). One point on a $10B network is our arithmetic. Corvane Airways, its rivals Northmere Airways and Aurelle Air, and every person named on this page are fictional — characters and sample data in the working solution.
Airline revenue management combines airline demand forecasting with seat inventory control: setting fare class availability, bid prices and overbooking per flight so each seat sells at the best price it can. Airline Revenue Management adds seven agents to that work: they watch every flight-date against its booking curve, explain each deviation with its sources, propose fare class and overbooking changes, re-check each proposal and review every decision after departure. The revenue management analyst decides, and nothing posts to inventory without her yes.
Bid price optimization starts from demand. The Bid price agent re-forecasts unconstrained demand for the flight, computes O&D displacement — the connecting requests a higher bid would turn away — and proposes a bid price and fare class closures with the estimated effect and spill risk. That is O&D revenue management, also called network revenue management: a seat is weighed against what every itinerary through it would pay, not only the local fare. The Checker re-computes it on its own. The analyst can ask why, change the number and approve.
Yes. The Deviation explainer writes one plain paragraph per flagged flight — a booking surge, a rival sellout, a school holiday, an aircraft change — and cites the bookings report, fare shop, event or schedule behind each clause. When it cannot find a cause, it says so and asks the analyst what she knows.
From show-ups and no-shows, not a yearly average. The Overbooking agent forecasts how many booked passengers will show up on that route and day — for example 97.6% of ORD–CDG passengers on holiday Saturdays — and proposes an overbooking level with the expected denied boardings and their compensation cost, which under 14 CFR 250.5 can reach $2,150 a passenger. It never goes above the cabin ceiling the director sets, and the analyst decides.
No. Agents may re-forecast, explain and propose; they never post to inventory, change a filed fare or a limit, or contact a rival. Every inventory change needs the analyst's signed yes with a reason code, and above her limits the Director of Revenue Management co-signs.
After each flight departs, the Outcome reviewer replays it with the old controls on the same demand and marks the decision helped, even or cost. It totals the results by reason code and suggests rule changes, which the director accepts or declines.
No. It reads your revenue management system's nightly forecasts and bid prices, the reservations and inventory system, fare shops, schedules, events and flown revenue, and works alongside them. Signed changes post through the inventory connection.
The Agentic Solution Engine builds and deploys it from your requirements and documents — your RM procedures, limits, reason codes and system connections — and it goes live once every quality gate has passed. First-time setup ends with a dry run on 120 departed flight-dates, where the agents propose and nothing posts, before a new portfolio goes live.
Every filing shopped before it publishes, so a mistake fare never reaches a shopper.
1-min demoCompetitive Fare MonitoringEach morning, the five rival moves worth money, with the captures that prove them.
1-min demoAirline Network PlanningEvery route tested on demand, feed, fleet, rivals and slots before it reaches the plan.
1-min demoWe'll run Airline Revenue Management on a sample of your own flight-dates, bookings and fare shops.