Hacker News new | comments | show | ask | jobs | submit login
Testing for common sense violation in airline pricing (arxiv.org)
47 points by kachnuv_ocasek on Sept 18, 2015 | hide | past | web | favorite | 33 comments

Some of these violations are common and ubiquitous. It's been well-known for decades that the UK train fares are not a metric, for a time even failing all three properties:

* Splitting tickets at an intermediate station sometimes reduced the total fare, so p(A,B)+p(B,C)<p(A,C);

* Reverse journeys are not always the same price, so p(A,B)!=p(B,A); and

* For a time it was necessary to purchase a platform ticket to enter the station proper, so p(A,A)!=0.

Even now I save thousands of pounds a year by finding optimal split ticket journeys, some of which take (partial) advantage of advance fare discounts.

So I'm not surprised, and no longer think of common sense as something to be expected from fare structures.

It may be more risky to try this with splitting your flights into individually booked segments.

Say, for example, I book a flight from ZRH to PRG with a stopover in VIE (which is usually cheaper than the ZRH VIE segement alone and backs the assertion of the paper up nicely, but I digress).

The legal connection minimum time in VIE is 30 minutes and Austrian Airlines will happily sell me a ticket with 35 minutes time to connect.

If the ZRH VIE flight is delayed and I miss my connection it's their responsibilty to resolve the issue. Potentially putting me up into a hotel overnight and even compensating me for a delayed arrival. EU rules are pretty strict here.

If I book the segments seperately it's my responsibilty and if I miss the connection due to delays I'm shit out of luck and the airline has no obligation or responsibility towards me.

I realize that the UK train network is very different than the scenario described and I appreciate that big savings can be made by planning ahead. I just don't think that this necessarily applies for flight ticketing.

Note that I'm not argueing that big savings can't be made by hidden city - or nested booking strategies. But those come with another bundle of potential problems.

Yes, that is indeed a downside, plus if you have hold luggage you have to consider the extra time to get it out of the first plane and checking it in to the second one which might well not be possible if the transfer time is as little as 30 minutes.

But assuming you have margin for that and since we are comparing the two same flights, we can assume that our loss is also the airliner loss, hence they will make everything they can to make the transfer work for the people that bought a single fare.

Therefore I see the difference in price as insurance against something that the airliner themselves have an incentive to avoid. Additionally to the risk, as with any insurance, I also consider the cost of the worst case scenario.

Most of the time my decision is not to pay the "insurance", unless time is critical and missing the connection means more than hotel + extra flight.

I've always found it interesting how round trips in one direction cost different than round trips in the other. I.E. p(A,B,A) != p(B,A,B). For example, I booked a trip October 2nd to Oct 17th from Singapore to San Francisco.

On United,

  $1170 - SFO UA 869,UA 895 -> SING UA 896, UA 862-> SFO
  $1690 - SING UA 896, UA 862 -> SFO UA 875, UA 7991 -> SING
I guess it makes sense that travel in one direction on certain days might be more significant than the other, But I find the $500 difference (almost 50% increase) to be pretty impressive.

I waited a few days, and re-ticketed again, and the price dropped from $1690 to $1463 - so that's also odd.

That's quite common within Europe, where a ticket from Warsaw to Zurich can be significantly cheaper than into the other direction.

In this case it makes a modicum of sense. Chalk it up to differences in purchasing power.

  I waited a few days, and re-ticketed again, and the price dropped from $1690 to $1463 - so that's also odd.
Not really. Airline pricing is very much based on mathematical models. When the system determines that the flight won't be sold for $1690 it lowers the price.

The trick is to fill every seat on the plane as expensively as possible, but not to leave any empty seats. Check out this article for details https://en.wikipedia.org/wiki/Yield_management .

In principle: There is no product going stale faster than an empty seat on a plane after the doors close.

Their goal is to maximize income, not fill seats. If it was the last flight they'd ever fly, you'd expect the price of an empty seat to converge to zero (or well, the cost of the fuel to lift a person and their baggage above the clouds) as the gates close, but you also have to keep the price expectation high for the customers silently haggling with the online ticketing system for a better price.

OT: it drives me crazy that yield management is not applied to events. If you have only sold half the tickets for your concert, reduce the remaining tickets. hn, do something!

If they did yield management properly at events which tend to sell out they'd also be able to capture ticket scalpers' potential profits rather than attempting overly elaborate ID schemes to prevent ticket resale. Not sure that many people would be thankful for higher ticket prices, but some of them might prefer that rather than having to buy tickets within an hour of them going on sale via a booking platform that struggles to cope with the traffic spike.

I used to work for a ticketing company. I know some Major League Baseball teams were making efforts at "dynamic pricing" which was basically yield management. Some ticket brokers do this as well.

> where a ticket from Warsaw to Zurich can be significantly cheaper than into the other direction. In this case it makes a modicum of sense. Chalk it up to differences in purchasing power.

Not just that. Everyone wants to be in Greece in August, not many people want to be in London.

Unless they're staying the whole month or the period is bracketed by major discontinuities, it isn't clear how that preference is relevant. A round trip is at least two flights. Why should the order of them matter?


It's pretty easy to imagine that demand could be quite different. For example, flying to San Francisco before the Super Bowl is going to be much more expensive than flying out of SF.

In the summer there is a daily direct flight from Hamburg to Whitehorse in the Yukon to cater to all the tourists pouring in. They flood to the Yukon in last spring, and leave again in the fall. The flight is somewhere around $$800 one way, $1500 return.

If you fly the reverse time of that i.e. Yukon->Hanburg in late spring and then Hamburg->Yukon in the fall, you can do return for ~$300 - because the planes are almost empty.

There's not really any good reason to expect arrival costs to be the same as departure costs for a given location. For instance, it takes much longer to board a plane than disembark, which means there's a comparative premium on leaving from less efficient airports. So even if it weren't about maximizing price, you'd expect different costs for different directions.

I believe airline ticket prices are often based seat demand and availability. SFO is United's west coast hub, so it could be that they are routing most trans-pacific flights through there and demand for seats is higher as a result.

Most airlines run giant data analysis operations on schedule (date and time) passenger utilization and anything else they can collect to derive an optimal pricing strategy.

Most fares are updated once per day and these days even intraday is possible. Most everyone uses SABRE, Amadeus and one other as their reservation backends which have various limitations (Spirit is an example that doesn't but it's very rare and theirs sucks). There are also a ton of caches everywhere to speed up the process which affects latency and increases stale pricing. It's sometimes very chaotic as to what everyone will see.

Also keep in mind that A->B and B->A and A->C->B, etc all have a different optimal pricing structure which is why you get so many odd prices. None of the reservation systems can consider every possibility either, they often have to short circuit a search or it would take forever so you can get variations depending on when/where/from what you look for these fares. Dallas to Austin is pretty trivial to search but small airport in Texas to small airport in China is ridiculously complex as there are trillions of combos.

Landing slots at airports have different fees. If you the A is a very expensive airport (e.g. Heathrow) the price of flying from and to A can be higher than the reverse.

The 2nd and 3rd there are not economically unreasonable, and the 3rd isn't metric-violating.

The 2nd would make sense if the net demand to go one way is greater than the other (think about moving trucks going from Detroit to San Francisco and vice versa).

The 3rd is the charge to go from outside the platform to inside it, not to stay in the same place. Whatever criticism you might make of the practice, the two A's are not the same.

Edit: And, FWIW, in case anyone hasn't heard the explanation, the "hidden city ticketing" in the first case exists when a provider with A-B routes and B-C routes wants to compete with a direct A-C provider and has to make the sum of the two legs (which takes more time) cost less than the direct, since it's necessarily less convenient, requiring them to vigorously scout for people who get off at B.

How do you find the optimal split?

Start by deciding on acceptable routes. Here's a graphic of a recent journey I computed:


Choose your date and time window for travel.

Then for every leg, download possible times and prices. Finally, simple search for the optimal combination.

You don't have contact details in your profile, but if you email me I can send you more details. I'm extremely busy just now, but if we establish contact I can send you relevant things in slow time. Let me know.

My experience is that the results from that site are very, very poor. My simplistic home-brew system regularly beats it by 30% or more.

"Common sense" tells us that air ticket pricing will be heavily optimized to maximize airline yield and that no effort is made to conform to any naive buyer conveniences or expectations. Although the data is interesting, the premise of this paper is a little weird.

Agreed, I think the assumption that ticket price is somehow related to cost is the violation here.

The article looks at common sense from the passenger point of view, but in fact if you look at it from an airline point of view you can sometimes see a very different picture.

For example, take the following 2 principles, each of which make sense individually:

1. A non-stop product from A to B is a superior to a one-stop product from A to B. Therefore, A to B non-stop is priced higher.

2. If only one airline has a non-stop product from A to B, they will price it higher than if multiple airlines have a non-stop product from A to B.

Let's see how this can play out in practice:

If many airlines can offer a one-stop product from A->B, your airline will offer it cheaply to compete. It may just so happen, as an implementation detail, that your product happens to connect at C since that's your hub.

Now, if you look at the A->C market, you may have the only non-stop product, and thus you price it high.

Result, your price for an itinerary (remember the actual itinerary is an implementation detail of the A->B one-stop product you purchased) from A->C->B is priced much lower than the person sitting next to you that is travelling from A->C only, even though the direct fuel cost to the airline of A->C->B is clearly higher than A->C only.

I'm curious how this was funded. If the authors had asked most anyone involved in Airline reservation management they would have told them this plain and simple as it is no secret. Yield management and airline network dynamics are just some of the drivers for these "common sense violations".

Also just to correct another comment on how often fares change. For international flights airlines can publish new fares to ATPCO up to 10 times per day however with newer pricing strategies/technologies such as bid-price the fares can change to meet the demand curve in near realtime.

Airline pricing and revenue management is a very complex and interesting topic that has always pushed the envelope of technology and business management. There is a reason why google purchased ITA. I was really excited to see a paper on the topic post on HN but in the end this study has very little substance to add to the many bodies of work out there.

Nice, now how can I use it to get a better fare?

I see that I should: * search for the round trip even if I go in just one direction * try to buy each connection separately if I have time to spare (maybe using multicity).

Any more tips? Is there a nice system that would automate it for me?

Here you go: https://skiplagged.com/

That site is based around booking flights with a connection at your destination and skipping the rest of the trip. Just remember not to check any bags!

Edit: It's called a "hidden city" fare.

Two other caveats:

Don't book a round-trip ticket like this, only one-way. The airline will usually cancel the return ticket if you skip part of the outgoing one.

Don't book on an airline where you have frequent-flyer miles you want to keep. In the past few years they're cracking down more by cancelling the accounts of people who use booking tricks that violate their policy.

> The airline will usually cancel the return ticket if you skip part of the outgoing one.

I recently did just that, buy a round-trip ticket with a stopover where on the way out I skipped the last leg. The return leg wasn't cancelled. I would've been furious if that happened, I paid for that and basically saved them money on the skipped leg.

I'd guess you either got lucky then, or it didn't look suspicious to them (e.g. your first time doing it, or there wasn't a huge fare difference from the ticket you "should've" bought). It used to be easy to use this trick, but airlines much more often cancel itineraries if you skip a portion now, since the contract terms specifically ban "hidden-city ticketing" and "throwaway ticketing". Was a bit of a brouhaha a few years ago at flyertalk when it started happening. But of course airlines don't disclose what exactly triggers enforcement.

"Testing for common sense violation" is the best thing I've read in a paper title in a while.

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact