I discovered those essays back in 2006 when I was working for an airline, and tried to get them to adapt ITA's technology for that reason. ITA was the best on the market at that time, and the e/web-marketing guys loved them too. But the bean counters deemed them too expensive and ruled them out. I left sometime after, was tired of trying to turn chicken shit into chicken salad (there were other issues involved, not just the data supply).
Does anyone know what makes ITA so special? Why don't any of the big airlines have APIs that Hipmunk, Kayak, Orbitz, etc. can use directly? (Do any of them?) Seems odd that everyone should have to rely on this one company for the data, when it should be originating from the airlines.
ITA's strength isn't data, it's algorithms. Finding the best flight is not just NP-complete, it's uncomputable. In practice, even a lot of sub problems are NP-complete.
People think the problem is a graph search problem of finding the shortest path. But that's actually the easiest part of the problem. It turns out that prices aren't attached to individual flights, but rather are rules of the form "If the passenger goes from A to B on a Tuesday, the price is X, and it doesn't really matter what flights they take, as long as the total distance is less than 3 times the line of sight distance."
Then there are all sorts of rules about what fares can be combined with other fares, under which conditions, ...
I worked at ITA software on QPX, and their code is by far the most algorithmically complex code I've ever seen. As I mentioned, there are many, many sub problems, and each one is a tricky algorithm design problem.
Orbitz and Kayak both use them, check the wikipedia page on them for more info in that area[1]. ITA's real claim to fame, at least in communities like HN, is that they're one of the largest Common Lisp shops still around. Also my experience interacting with their employees has been that they're an exceedingly smart and talent bunch of people. Additionally, and especially if you lived in the Boston/Cambridge area, they're pretty well known for there rather challenging interview puzzles which used to be plastered all over the T [2]
The thing to understand is that the airlines publish data that ITA uses. The problem is that it takes an incredible amount of engineering effort and algorithm smarts to go from 'published airline data' to 'this is the cheapest flight from BOS to LAX on 4/20'. ITA does have competitors; historically those competitors offered much lower quality products, although that might have changed.
ITA is special not because of the data but because of what it does with the data. The others haven't been able to reproduce that. Think of it as the Google of airline data.
The data is available (not cheaply) from an company called ATPCO, which is owned by the airlines. The airlines publish fares and rules through ATPCO, and ITA, SABRE, and many others fetch the data and run the computations.
So if Hipmunk, Kayak, or whoever wanted to remove their dependency from ITA (by doing more computational stuff themselves), what's the best place for them to get their data?
I'm pretty sure ATPCO is the only way to get the raw data; people who subscribe to it are not allowed to redistribute it.
As for getting processed data, that's called metasearch: scraping other sites' search engines. Kayak started out as pure metasearch, but moved to using ITA. Not sure whether they still do any scraping.
While your thesis may be correct, some of your assumptions are a bit off:
1) Bing has stopped investing in travel tech, they now use Kayak for search. So do we still believe in the value prop?
2) Farecast (fare predictor) turns out to be not very valuable. So much so, that Bing couldn't even sustain development of their own search tool. Can Google do a better job of it? Maybe. Can they do a good enough job of if that it's ROI positive? I'm skeptical.
I agree that Google believes many of your assertions. I don't agree that this makes it true. Do people really need more sophisticated travel search functionality? Sure, you can imagine lots of cool stuff -- bells and whistles. But ask Kayak if their "Explore" feature (perhaps a lite version of what you can imagine Google doing with ITA) drives their business? I'd argue it's neat, good for branding and loyalty, but not at the end of the day, a major source of value.
That being said -- the key thesis of: Google trying to hold onto relevance in travel search -- is true. However, they are still pretty much a one-horse show (search revenue), and travel is one of (if not the) biggest vertical today.
Right because before I had to keep random acquaintances as friends and read their updates in order not to hurt their feelings, now I have to fly to random places in order not to hurt peoples feelings. Can't wait.
I had a business acquaintance get her feelings hurt because I didn't follow her back when she followed me on Twitter. I looked at her feed and she literally had tweeted about her breakfast that morning.
I quit Twitter instead so that I wouldn't have to hurt anyone's feelings anymore.
I wasn't getting much value out of Twitter anyway (not a judgement on Twitter's value, just on what I was getting out of it), and my business depends a lot on maintaining healthy relationships with lots of people, since we grow primarily by referral. It just made sense to drop Twitter altogether and tell people instead that I don't use it.
I don't think that's a problem for everyone (I aim for linkedin/facebook separation for these reasons)... Nevertheless, you could filter by the "hide this post from" list, weights of friendship (amount of likes, views, etc.) and other statistics... Of course, this is only viable if you have friends/family all around the place, not just in your locality (definitely not for everyone).
Still, seems like a nice advertising focus trick.
I'm working on something like that -- I'm starting with discovering where you can go on your airline miles, but it's all visual. I'm planning on integrating suggestions of low miles tickets where your friends are. I know this isn't quite what you're suggesting, but I'll be getting there :)
I just have a launch page up right now: http://flybymiles.com - but private beta coming next week.
In the meantime check out http://wanderfly.com - they could use a much better recommendation engine, but the content is there (ie: beach for $500)
I'm sort of surprised this went through since their advantage in search mixed with the power ITA has within their field leads to very powerful mix that could put competitors at a serious disadvantage. Maybe the DoJ saw something I did not to make it less of a problem, but even as a google fan it's... ungh. I don't like this going through.
The key was Google's agreement to continue licensing QPX on “fair, reasonable and non-discriminatory terms”. According to the linked Justice Department press release, they're also required to continue development of the publicly-available version of QPX.
Very interesting. Be very curious to see if it is enforced (although I think google is more prone to not risk it since they are under so much scrutiny if nothing else).
So, I have an interesting perspective to share. I'm graduating and might end up working for Kayak. I know at least part of their infrastructure depends on ITA, and while the existing contract goes until 2013 (I think), Google is obviously looking to compete. Kayak also recently filed for IPO.
It's not like Google has never failed at any of its ventures, but to me it seems like it might be a tough competitor to everyone in this field, since a decent amount of traffic does come through search results. On the other hand, I think a majority of users come in directly, either through WOM or advertising or previous use.
So, what is HN's thoughts on the situation for Kayak? Would it be a wrong move to accept a job there? The engineers seem smart, the work is interesting, the culture is sane and easygoing. What's the worst that could happen?
Kayak is a great company and this is a fascinating time to be in travel search. Go for it! A first job under Paul English (even a few layers beneath him) is a big win and Kayak's engineering reputation so great that you'll have plenty of opportunities down the road.
So in summary: Worst scenario is pretty darn good for you!
I wouldn't discount a company because they're competing with Google. There are a lot of companies that have competed with Google and have successfully out-competed them (likewise, there are a lot of companies that Google has acquired after failing to out-compete). You should look at Kayak on its own merits -- if the folks there are giving you a good offer, and you think the company has a good chance, then go for it.
I believe the consent decree was reported to say that ITA has to renew existing contracts until at least 2016. So that's 5 years before Kayak might, theoretically, lose access to QPX. Don't worry about it; 5 years of a good job is a good run.
This is a good opportunity for someone to start a competitor of ITA. Google are only required to serve the ITA clients for five years. They will probably stop afterwards, they will want people to use their search engine to find flights.
The only thing special about ITA is their excellent software. Apparently they get their raw data from another company that is willing to license the data to anyone. I am sure that there are still some old LISPers around that can write a similar software for a new company.
Apparently they get their raw data from another company that is willing to license the data to anyone.
This is not true. In order to do what ITA does, you need schedules, fares, and availability data. You can buy the first two from clearinghouses if you have enough money. But the only way to get availability data is to either contract with airlines directly OR contract with your competitors.
I am sure that there are still some old LISPers around that can write a similar software for a new company.
Breathtakingly wrong? Are you saying there aren't any old lispers left? Or that old lipers cannot write similar software?
Before you say something is breathtakingly wrong you should read it carefully and try to understand it. This way you might also get some of the jokes that are currently flying over your head.
There are plenty of lispers left. I'm one of them. I also used to work at ITA.
The bit that is breathtakingly wrong is the notion that you can just take some smart lisper, lock them in a room, and have them replicate what ITA did and have that be a viable business. See martincmartin's and my comments later in this thread to learn why that would be incredibly hard. With a smart enough team and enough time and money, you could certainly do it, but this is a project that would take many years and its unlikely anyone would be able to keep funding for that long. ITA had a much easier job: it was competing with very poorly designed systems that ran on mainframes and had very little CPU power available.
Given what you and martincmartin are saying, I am left wondering why Google was able to buy ITA for so little. How can such a thing be worth less than a billion dollars?
Because there's a rather finite amount of money spent on air travel which means that the valuation of any one company in that sector will be well within Google's means?
Perhaps. But in that case it's a much smaller market than I imagined. Also relevant is that ITA raised $100M in 2006, so their $700M exit is hardly a home run by VC standards. It doesn't seem to match the astonishingly strong technical and market position you (convincingly) describe.
ITA raised all that money to build a reservation system to replace the old mainframe systems that are fabulously expensive to run. That would have been a market of the kind that required outside investement from a company that was already profitable (I think ITA had more people working on the reservation system than the search), and that a big VC would love to bet on.
From what I hear it worked technically, but failed commercially due to the badly timed recession. It seems like a fair bet that Google's offer wasn't putting a lot of value on the reservation system. Even if it could be made to work, it's not the kind of a business that Google is usually in.
The margins in air travel are also slim. Google did a very smart thing here. There business model can support the ITA purchase, not many other businesses could do this. The kicker is they also just hired a lot of talent.
It would be good to read up on ATPCO. Basically, ITA sucks up the ATPCO feed as well as any direct connects with airlines worldwide and then processes that data into usable feeds for clients. In many cases, the trouble/expense of handling the ATPCO data is not worth it to sites so they pay ITA to provide them with a feed that meets their responsiveness/API/uptime/accuracy requirements.
How? They already have the data and plumbing, all they need is a GUI. I don't think Google would go for a Hipmunk-style GUI, they'd just build something search-based you could see in your results.
It shows Google wants to be in this space and is willing to buy into it. I don't know that Google will specifically go for Hipmunk, but when any major corporation starts buying into a market, they tend to make more than one acquisition, and it's often the partners of the big companies they acquire first. I'm pretty certain they'll make a couple more strategic acquisitions in the airline/travel space, and Hipmunk is one of the more interesting partners of ITA, since they're new enough to probably be cheap (single or double digit millions, if Steve were looking to be acquired), and all the original team is still in place, no one has cashed out. Hipmunk is also one of the most interesting teams in the space.
Further, if I were doing a talent acquisition, I can't think of many people I'd want more than the reddit team, and Hipmunk is currently pretty much made up of the early reddit team. Of course, Steve is already rich (by some definition of rich), so probably isn't desperate to sell for merely "talent acquisition" money, but Google can afford to stretch the meaning of talent acquisition, and have done so in the past to get really interesting people.
Also, saying "all they need is a GUI" is vastly underrating the importance of the GUI. I can imagine an Android app with a Hipmunk style UI being really powerful. Sometimes you want to find the perfect balance of price and schedule, and sometimes you want to do that while in the middle of a meeting or in the cab on your way to your hotel so you know when you're flying again, or whatever.
And, to get specific about YC-related acquisitions, Google docs had a spreadsheet when they acquired Zenter. All they needed was a GUI, but they acquired a two-man company that had a few tens of thousands of lines of code. Probably a talent acquisition (and a smart one, since Robby and Wayne are awesome), but nonetheless, an acquisition.
Anyway, I agree that Google will definitely be building this into the search user interface. But, when Google moves into an area they don't just do the search interface. Maps has loads of entry-points into it, including non-search oriented use cases, and Google made several acquisitions and developed a bunch of stuff internally to make Maps happen.
It could also be viewed as epic validation that the ideas are gold. I most certainly believe Hipmunk can compete against Google; the team is completely awesome and they're far more nimble and agile. Although Google does have an unbeatable vision of Maps I'd imagine...there's no real reason you shouldn't be able to get bus times and airplane times in the same place. And perhaps if you could then pay for the bus and the plane using your Android NFC...
They could probably spin new features faster than Google. For example, every service I've seen requires a departure date and return date. In my head, I thinking, "I don't know - sometime this summer." I'd like to be able to pick a destination, the number of days, and have it tell me when I should go on vacation to get the cheapest fares.
rough times ahead for flight and travel sites (which generated good money via leeds to flight search sites) ... googles killing one internet ecosystem after the other.
In brief:
- Google will be required to license QPX software to airfare websites on commercially reasonable terms.
- Google will be required to continue to fund research and develop the product “at least” at similar levels to what ITA had invested.
- Google will be required to develop and offer ITA’s InstaSearch product to other travel websites.