Yes they do use stochastic optimization. The terms to search for are "yield management" or "revenue management" and are often used by airlines and hotels. To a lesser extent retailers have applied the methods to optimizing discounting on seasonal goods.The airline version is especially interesting because: 1) the price-sensitive demand arrives before the price-insensitive demand; 2) the system features signaling by customers (e.g., "search by schedule" vs. "search by price/flex-date" modes in booking engines); 3) competitor dynamics (price transparency with third-party booking engines) 4) extremely high cardinality (making {pricing x capacity} decisions on billions or trillions of {Time_Departure} x {origin} x {destination} x {Time_Return} tuples)

American Airlines (AMR) were a pioneer in the use of operations research to set ticket pricesThey span the group off as Sabre>>unquestionably the biggest and arguably the best and most influential OR group in the worldSabre Soars

I've not modeled this myself although I've been to a couple of industry conferences that discussed these models. It's all part of a long-term book project of mine on organizational microeconomics in networks with distributions of utility functions.By cardinality of {time}x{place}x{place}x{time}, I mean that ALL four variables do impact the price offered to the customer. Airlines absolutely DO NOT price flights independently of each other (although booking engines may display prices as if that were true). For example consider the pricing of a round-trip between Amsterdam and New York. Someone that asks for week day departure with a return 24 hours later will get an extremely high price because they are mostly like a business traveller. Someone asking for the same week-day departure returning two weeks later (a holiday maker) will get a much lower price. Time of day and day of week on both ends of the round-trip will impact the price. The cardinality may even be higher than this 4-tuple suggests because of route combinations of non-stop versus various connecting-flight journeys -- e.g., a business traveller may seek the fastest non-stop flight (high price!) while a holiday maker might pick flights with one or more changes of plane (low price). The point is that the problem is not separable although some airlines reduce the cardinality by defining categories of round-trip structures (e.g., Saturday night stay or no Saturday night stay or use heuristics based on day-of-week). The span of ticket prices paid for passengers on the same flight can easily exceed 10:1 although the transparency of prices and rise of low-cost carriers has probably attenuated this. Overbooking is subtle game. I know of one airline in which the "cost" of bumping a passenger to the next flight is less than 1/2 the same-day walk-up fare. The airline can easily clear more than $400 in pure profit by selling a high-price ticket on an overbooked flight and offering volunteers a "free" flight (with strings attached) in exchange for waiting a few hours for the next flight.

That article was very interesting. I've never personally heard of operational research. I have had a fair amount of experience working with AI solutions to problems, and his claim that AI offers no competition against OR is an interesting claim.Has anyone here worked with OR to solve problems related to market decisions? I know people try to use AI methods all the time, with what seems to me limited success. It would be interesting to learn about operational research solutions for market problems.-Rob

QuoteI know people try to use AI methods all the time, with what seems to me limited success. It would be interesting to learn about operational research solutions for market problems.AI was the Internet bubble of the late 80's. It has not arrived. The best that emerged in the early 90's were expert systems which downgaded to rule-based systems. It was a disappointing era. The fact that you drive in a reliable auto is due to OR. I think the origins are from WWII (logistics). You need graph theory for OR (use the Boost Graph Library)!

OR is good magic -- providing a wide range of approximate and exact algorithms for constrained combinatorial optimization. Some claim that Wal-Mart did more to prevent inflation that Greenspan did and Wal-Mart's secret is OR. The retailer hires a fair number of Ph.Ds and maintains a 583 TB datawarehouse that used for every-day decisions on how much of each of over 100,000 different products to carry in each of about 6,200 stores.I'm not surprised that OR would be considered better that AI for a problem like airline yield management. Although the yield management problem is complex, it's very structured. Moreover, the feedback on "learning" is too long and the causal structure is too nonstationary to allow AI to operate to best effect. It would be hard to make an AI that generalizes to events that change price elasticity, booking arrival patterns, or no-show rates (e.g, 9/11, a competitor's bankruptcy, or the entry of a low-cost airline, or the rise of online booking firms).

FWIW, Columbia's FE and OR programs fall under the same umbrella Columbia IEOR.

QuoteOriginally posted by: Traden4Alpha Some claim that Wal-Mart did more to prevent inflation that Greenspan did and Wal-Mart's secret is OR.that, and selling crap made in china that costs a fraction of what american made goods used to

QuoteOriginally posted by: CuchulainnQuoteI know people try to use AI methods all the time, with what seems to me limited success. It would be interesting to learn about operational research solutions for market problems.AI was the Internet bubble of the late 80's. It has not arrived. The best that emerged in the early 90's were expert systems which downgaded to rule-based systems. It was a disappointing era. The fact that you drive in a reliable auto is due to OR. I think the origins are from WWII (logistics). You need graph theory for OR (use the Boost Graph Library)!can you elaborate your points?

QuoteOriginally posted by: JQuoteOriginally posted by: CuchulainnQuoteI know people try to use AI methods all the time, with what seems to me limited success. It would be interesting to learn about operational research solutions for market problems.AI was the Internet bubble of the late 80's. It has not arrived. The best that emerged in the early 90's were expert systems which downgaded to rule-based systems. It was a disappointing era. The fact that you drive in a reliable auto is due to OR. I think the origins are from WWII (logistics). You need graph theory for OR (use the Boost Graph Library)!can you elaborate your points?Which ones in particular?

QuoteOriginally posted by: ppauperQuoteOriginally posted by: Traden4Alpha Some claim that Wal-Mart did more to prevent inflation that Greenspan did and Wal-Mart's secret is OR.that, and selling crap made in china that costs a fraction of what american made goods used toRight, because we all know Wal-Mart is the only retailer selling Chinese goods.