As a co-founder of ITA Software (travel search) and founder of the company that makes http://inky.com (email), I often see unexpected parallels between travel search and email. In both spaces we get periodic design documents and slide ware that look really cool and get everyone all excited. But in both spaces these designs rarely get implemented, much less reach production.
The reason is that the domain details are so extensive and difficult that almost nobody can get past them. I like to describe both these problem domains as "fractal," because when you're 100,000 feet up it looks pretty simple, but the closer you get the more details there are. If I had a dollar for every hacker who told me how easy it would be to make a travel search web site or an email client, I'd be rich.
In both spaces, a very small number of players create the core technologies, and a much larger set of players layer on top of these cores. In travel, for example, you have the GDS companies (Sabre, Amadeus, etc.), ITA, and Expedia. Everybody else -- literally, everybody else -- layers on top of one these systems. Kayak? ITA customer. Hipmunk? ITA customer. Etc. This means the vast majority of players don't actually have to deal with the fractal domain complexity. And almost nobody has any clue there's any difference between say, an ITA and a Hipmunk, even though one has a million lines of code and runs thousands of machines and the other has a fairly standard website. (#)
Similarly, in email, a handful of players create real mail stacks. These are the usual suspects (Microsoft, Google, Apple, IBM) and a handful of others (Inky, Sparrow, Thunderbird) Everybody else -- literally everybody else -- layers on top of GMail or Outlook. Mailbox.app? Layer on top of GMail. Xobni? Layer on top of Outlook. Postbox? Fork of Thunderbird.
In the mail space, there are at least some decent open source libraries to use. In the travel search space there is literally nothing to start from but a blank sheet of paper. Carl de Marcken did most of the early work figuring it out for ITA, and it nearly killed him.
I actually seek out domains like this; either because they're defensible, or because I'm insane. I'm not sure which yet.
(#) If you're thinking it's necessary or even better to own the million lines of code and run thousands of machines, think again: Kayak created substantially more value for its shareholders than ITA did, with a lot less effort.
I imagine it's a similar problem to online payments. From 100,000 feet payments looks like shuffling bits around and keeping track of them. And in some sense it is. But in reality that technical problem is maybe only 1% of the business, the rest is things like customer service, fraud protection, auditing, regulatory compliance, relationships and integration with other financial companies, relationships with governments and law enforcement organizations, etc. Each of which is a hugely different problem and vastly more difficult. The funny part is if you translated the problem to the physical realm few people would fool themselves into thinking that a bank could be reduced to little more than a ledger.
I actually disagree. I think getting users is a subproblem of creating a MVP (Minimally Viable Product), which is incredibly hard in these spaces. You can't get users unless you have an MVP. But what if your MVP takes 1000 person years to create? (That's not the case for email or a travel web site, necessarily, but certainly is the case for an airline reservation system... or a web browser.)
Remember Vonage? All they had to do was offer an MVP for voice phone. But it turns out that if a customer picks up the phone once a month and doesn't get a dial tone and a clear connection, they're going to abandon you.
The Lean Startup wisdom about focusing on an MVP and getting to market fit is absolutely right. The issue is: what exactly is the MVP for your space? Not all MVPs are equally easy to create. Try creating an MVP for a nuclear reactor, for example. Well understood tech? Yes. Easy? No.
<a href="mailto:mail.added@by.js" />
some
<span style="display: none" />☃</span>
person@gmail.com
</a>
And here I can either click the link or copy the text and it isn't all messed up. A simple function can easily generate that HTML for any email string.
You're being downvoted with no explanation. The issue with that is that your email address is easily harvested by combing through websites. The comment above yours is describing a way to get around this by obfuscating the email address in HTML and using CSS trickery to get it to show normally.
Clicking that link to inky.com, I had to roll my chair 2 meters backwards to view that screenshot without getting the feeling it's pushed into my face. Large fonts and design is great, really, but in my personal opinion this is too large.
Edit: And I hear from others that they can actually download it. I was looking for "where the hell is their call to action, can I test it somewhere?" but there is just nothing. Turns out that Windows users do get a download link, and Linux users are told "shut up you are not supported and we have no clue that humans invented emulation software that might just work, or at least tell them they can try it in a Windows VM".
You have a point. The question for us is: do we want to promote Inky under Wine as a supported platform? Since we don't run it that way ourselves (ever), we're not terribly comfortable doing that.
We do have the core running under Linux and are working on getting the UI up under Linux. We'd really like to have Linux Inky for ourselves, if nothing else, because Linux offers the best environment for dev/debugging. So it's coming. :)
> In travel, for example, you have the GDS companies (Sabre, Amadeus, etc.), ITA, and Expedia. Everybody else -- literally, everybody else -- layers on top of one these systems.
I thought Southwest was known for having their own system. Under the hood are they actually using a Sabre, Amadeus, etc.?
I actually seek out domains like this; either because they're defensible, or because I'm insane. I'm not sure which yet.
I see this quite a bit. In fact, I strongly believe there are only two types of businesses that make competition improbable: A) those that pick normal problems, show up to the market early, and price to discourage competition, forever locked into low margins, and B) those that choose incomprehensibly hard problems and solve them better than anyone else.
So do you see this concept as a better or worse Kayak? I feel that google has done the best job with flights and they are my go-to location. This concept seems great for trips, but I'm not sure how many people per year are flying for a vacation vs flying for necessity.
There are lots of great ideas here, and the design is lovely. But my guess is that many of the design concepts would disintegrate when intersected with the practical considerations of actual airline data and systems.
Totally random side note: my friend Jason Rubin (co-founder of Naughty Dog, which I worked for in the mid 90s) suggested the map-with-flight-path-arcs representation when I showed him prototypes of QPX, ITA's travel planning system, circa 2000. I like the idea visually, but in practice it gets messy (think around-the-world trips).
Yeah I didn't really think of the around the world use case. At that point, you'd need to change the visualization to a globe which would probably confuse most travelers.
Terrible idea for selling airline tickets [1], but perhaps less terrible for selling design consulting.
[1] The airlines believe that, overwhelmingly, their customers know where they're going and when, and care mostly about how much. "Enticing maps" and "impressive photography" are unlikely to increase conversion/task success rates versus typing in "LAX"/"NGO."
The silky smooth transitions are nice, but unfortunately no amount of front-end UX work will make the backend not take several seconds to look for possible routes/pricing for you. The multi-page workflows for e.g. Delta.com (which actually don't suck) partly help to obscure how dog slow the backend is relative to Internet Speed (TM). A successful rework which made the app feel much more responsive could have many customers offer the feedback "THE SITE IS MUCH SLOWER. WTF." and the fact that this feedback is objectively untrue would not prevent it from costing the airline hundreds of millions of dollars.
In fairness, obfuscating the slowness of the backend system is one area where fancy transitions involving planes flying across maps and irrelevant trivia like weather symbols and "social proof" can actually potentially help, providing it's interesting enough for people to actually pay attention to.
Fancy graphics are unlikely to sell many flights, but I can't help wondering whether better visuals added at the right stage of the booking process might help with ancillary revenues from hotels (and onward flights with one way tickets). Maps, for example, are a pretty horrendous way to pick the primary destination, but a potentially useful device for highlighting connecting flights between long haul hubs and leisure destinations (bet you didn't know you can book through to Phuket with us?) The interesting work there is still little to do with UI and a lot more to do with getting the back end to spit out relevant suggestions without bringing everything to a grinding halt.
Then again, the airline that is perhaps most dependent on ancillary revenues generated by web upsell is Ryanair and their website is purposefully ugly for branding reasons (ugly = the cheapest) as well as deliberately confusing for improved conversion for add-ons.
I believe patio11 is positing -- I would -- that the user already knows his destination (or has a good enough idea) before waiting on a backend search for pricing. Maps, trivia, weather, etc are useless information in the majority of cases. Transitions appear useless. Just load my pricing data already! Stop wasting time loading this crap! It's like 4MB Flash, splash screens before I can load your 100K site. Yes, technically, in the airlines case, we can probably load a static image and trivia in the time it takes to query the backend, but the users are not technical and it just has the appearance of fluff, and they just want prices!
It would actually be interesting to test cases where a user would rather look at a loading screen than an image attempting to distract from the loading or vice versa.
Either way my main point was that well designed and tested apparent fluff that might lead to a tiny fraction of users abandoning searches before getting a price could make up for it by encouraging [other] purchases. If there's a time lag you may as well advertise in it.
Consumers running exploratory searches direct on an airline's website rather than via a comparison engine you might not be that price sensitive anyway...
What? No. Why would you put a weird palm tree icon on a marker instead of the price? I know it's warm in the Mediterranean, I'm not an idiot so please don't treat me like one. Also, why does it matter where the user is located and what the temperature is there? I know where I am (which, in many cases is different from where I plan to fly out of) and I can probably figure out the local temperature if I need to, which I don't when booking a flight.
Also, "icon-driven" navigation? In other words, confusing pictures that I have to think about instead of clear text. Drop the icons and the interface actually gets more minimalist (since the text is there anyway).
How in the world did design get to this point? It's like somebody decided that the Internet was too convenient and efficient and asked a bunch of designers to figure out ways to make it harder to use instead of fixing the actual problems that exist... (an example of which would be the date dropdowns on travel web sites, let me either select a date or type in a free-form date that gets parsed smartly).
That really isn't the type of input I was thinking of, though certainly that is a fairly easy problem to solve in a variety of ways. I'm more interested in being able to say something like 'Aug 2' or 'next Thursday', the way I can in Google Calendar.
Oh hey, that does work! I hadn't noticed that, good for them! Their calendar dropdown is also superior to most others in that it isn't a popover-style context menu but has been integrated into the interface so that I can see a visual representation of the dates I've chosen.
Now if only their site wasn't completely broken on mobile... :(
I fly twice a year. Why the hell would I want an app for a flight booking company? It's this sort of mentality that really gets my goat when it comes to mobile.
I don't get how this is non-logical compared to what I said earlier. 24hour is the same excuse as the empirical system, you can divide 12 by 2,3,4,6 thus it makes things easier. It's a different system than the metric one, we could have said that a day lasted 100hours and week 10days. It doesn't make it non-logical though, it's like changing the basis.
The original comment I replied to was suggesting that the date picker was a bad thing. My point is that it's better than any other method, since parsing a date correctly for all people in all situations in the way implied in that post is not possible. There will always be ambiguity over the date if the user chooses to use 'dd/mm/yyyy' or 'mm/dd/yyyy'. Chaking location, or locale isn't much use - we brits go to America on occasion, and they come over here. If you mark it as [dd/mm/yyyy] then you've just limited the usefulness of 'a free-form date that gets parsed smartly'.
You would handle that by looking at looking at a user's location settings and choosing the appropriate one. For example, if we detect your IP is in Germany, we would prompt "Would you like to see this site in German and change the currency to Euros?" If you say yes, then we would also localize the date formats.
> You would handle that by looking at looking at a user's location settings and choosing the appropriate one.
Great, now if I try to book a flight from Spain to Hungary while being in the US for some reason, I'm pretty much guaranteed to get the one date format which makes no sense whatsoever.
If you were in the US, we would use US conventions unless you had specified that you picked a different localized site (let's say the .es site) in which case we would switch to the conventions based on localization settings.
Did your browser ask for en_GB or en_US or fr_FR or whichever localization? What is your IP? With those two info, I can make an educated guess and print out the expected format mask near the input box.
No. I work for a travel company and not only does this not work but it is likely illegal under DOT in the US. Using a map to pick is cutesy but pointless as you might not even know where the hell you are going except by name or airport code. I admit picking flights is rarely much fun except at sites like Google and Hipmunk who do not book and are thus not limited by the DOT and a wall of lawyers. Also GDS systems (and those airlines that handle their own reservations) are slow as shit to do anything. Google manages to be fast because it caches heavily producing really stale pricing and availability which is not easy to overcome. You can have fast or you can have up to date but you can't have both. Plus you have all the issues that everyone in the industry hates everyone else and thus makes interoperability a pain in the ass.
Preference-based display of flights (I'm thinking).
Once upon a time, AA owned SABRE and all the flight information was in the one GDS (Global Distribution System). Then somebody at SABRE got the bright idea to preference the order so AA flights were always on the first page and all other airlines were on the second page. Most bookings were made from the first page.
This was then made illegal and they said some gobbledy-gook like: "you can't preference the display, you have to show a reasonable number of flights, all in the same format, and sorted by time, price or duration".
Or something. I can't find a reference to it, but it had to do with SABRE's virtual monopoly on global travel distribution (inventory, agents, etc) and regulation / de-regulation of the travel industry.
The DOT has tons of rules over how you show prices and other data about flights. They want the customer to know everything about what they are about to choose and make it difficult for anyone to sneak something in. Plus our lawyers are so scared of fines they go overboard to follow every exact rule. The DOT also loves to change the rules or their interpretation whenever they feel like it so it's a moving target. You can make it perfect then they change the rules and you get fined anyway. This means anyone who does booking is always on the lookout for potential fine issues and this makes innovation impossible. If you don't book the rules are not as bad.
For example if we fail to show that a flight on one carrier is operated by a different one (very common for some routes) they can fine us tens of thousands for each one we get wrong. A very strong incentive to be careful!
> Using a map to pick is cutesy but pointless as you might not even know where the hell you are going except by name or airport code
YES. Just trying to pick my time zone on the OS X Preference Pane is a exercise in frustration, I end up typing it in 70% of the time. I can't even imagine when you want to pick a specific airport on the other side of the planet.
I agree, but if this were for an airline as it proposes to be, at least they would presumably be able to access their own ticket prices quickly, or could upgrade their systems such that ticket prices can be available more quickly. Also, they show a map of Europe, so presumably U.S. Department of Transportation regulations wouldn't apply.
Ah, I've noticed that with Google. Very annoying when the price is wrong. Although it bothers me even more when I'm on a site like southwest and it makes me restart because a price has expired while I'm buying it.
It's not hard to create beautiful layouts and cool interactions when one is not constrained by real-world limitations, restrictions, users, budget, and technologies.
In general I'm not a fan of this sort of splashy, spec or concept project. I think it telegraphs that a company is not focused enough on real work, or at least that they do not understand the crucial importance of constraints to good design.
It reminds me of this rant about tech company "concept videos", which I very much agree with (ignoring if possible the MS vs Apple minefield):
It additionally reminds me of this particular agency's (f-i) previous HN submissions and tagline plastered on all their pages ('we built the new USA Today site, IMHO
I hope not, that UI is infuriating to use. I know it isn't exciting or glamorous, but the type-ahead dropdown thing everyone uses right now is used for a reason- it works.
Even worse, this concept focuses entirely on airline websites for booking travel. I think airlines need to work a lot harder on what happens after that- the ability to easily look up baggage limits, airport info, checking in online... all of these things are a lot more important than booking, which most people don't even do on an airline's own site anyway.
This is what I want when I'm looking to plan a trip. Notice I said "trip", not flight. Usually I don't care about my method of travel, I just want to get from A to B to C in an efficient and affordable manner.
I live in Central NJ. Nearby me are several Amtrak and NJ Transit stations and three major airports (EWR, JFK, and LGA). I have a car that can take me to any of those places. I also have friedns that could drop me off at any of those places as well. Plenty of options to depart from.
Let's say I want to visit some family in Florida. My final destination is also abundant in transportation options. For Florida, I can land in Miami (MIA) or Ft. Lauderdale (FLL) and be picked up in either one.
I usually end up fiddling around in Excel, planning out time and costs for many different ways to get from A to B..
EWR -> MIA
EWR -> FLL
JFK -> MIA
FJK -> FLL
... etc ..
And to throw another wrench in the plan: dates. I could be planning a trip a few months in advance (in hope that tickets are cheaper). Some sites really struggle with the concept of having flexible travel dates. I want to be able to say:
"I want to leave from the general $HomeArea,
go to $FinalDestination, +/- $x days of $date1,
and come back +/- $y days of $date2."
I shouldn't be building my own OLAP cubes in Excel to not get screwed over on rates.
But I've grown enough despise towards the travel industry that I'll happily spend a few hours to make sure that I'm not paying them a penny more than I have to.
ITA Matrix is awesome; sad thing it doesn't seem to have data on cheap airlines that operate in Europe (Ryanair, Easyjet, WizzAir, et al.), as they tend to be used here most frequently.
I usually end up fiddling around in Excel, planning out time and costs for many different ways to get from A to B..
EWR -> MIA
EWR -> FLL
JFK -> MIA
FJK -> FLL
... etc ..
What are you talking about? Every major travel website has an option along the lines of 'New York City - All airports', which includes JFK, LGA, EWR
Flexible dates are a nice thing to have but I wonder how much they would actually be used? I cannot think of a time where I was booking a flight for business and not only were the departure and return dates non-optional but I was looking for pretty specific times as well (i.e. within a few hours time range).
Even when traveling for pleasure I generally want specific dates so as to maximize my time at my destination.
The problem is that each person has different needs and it's difficult to make a site that is usable for most of them - that's why so many 'niche' sites exist.
For example I used to fly a lot and most of the time I did not care about exact dates at all. I guess that scenarios like 'I live in Spain now and I want to visit my parents in the Czech Republic some time in June' or 'I want to spend some weekend in Paris' are pretty common.
Sometimes I don't even care about destination that much - I want to travel somewhere in south of Europe to take a couple of days off - all I care about is the price. Date-of-flight-centric sites are really difficult to use for me. Only recently I discovered http://skypicker.com (I'm not affiliated wit them) that does what I want - I want to fly to Spain some time in August - here are cheapest flights.
Skyscanner and Momondo offer that to some extent. You can choose departure/comeback time as "whole month" and airports as sth like "London (all)". But you can't select say two different outbound airports from different cities.
NYC works for most sites. I wish there was an SF equivalent. I usually don't care SFO/OAK/SJC and have to do what you do, although I just keep it in my head.
Why? This is a pretty site that has a bunch of usability flaws making it pretty useless for actually searching for flights. And all the other reasons mentioned in this thread…
As a very frequent flier, I want more power tools, not prettier interfaces. When I had the status on Delta, I ended up booking most of my tickets through the Diamond Desk simply because they had tools that let us much more efficiently search the itineraries I was interested in. For any given trip, there are often many thousands of ways to fly it, with very complex trade-offs between the options. Once you get to multi-city trips with things like open jaws and stopovers, all hope is pretty much lost if you're trying to do it without working with the airline.
What I really want is a text query interface to QPX without needing to be an airline and pay seven figures for it. (That's not hiding somewhere in the Matrix interface, is it?) Especially if I could cross-reference it with flightaware's revenue and load data, which is super useful for things like predicting upgrades. Kayak's nice, but it still takes hours to go through all the options, flipping between the tabs. I don't really get Hipmunk.
TL;DR: Flight search is extremely complicated and these screen shots don't cover any of the things I find difficult.
In fact, I often use Kayak to figure out which airport I should fly into. For instance, if I'm traveling through countries X, Y, and Z and I don't care what order I visit them in, I can quickly check which one is the cheapest to fly into.
Search interfaces seems to be stuck in the 90's too.
There are just so many problems with travel sites you can't possibly say it is mainly an UX issue (well, you can since this is your business). What's really soul-draining is spending hours looking for something, finding the perfect thing (or surrender to the less awful one), filling a ton of information, put your credit card in and be told that an 'unexpected error' or 'could not find your flight'. Another thing is to be instantly notified of promotions, I have just yesterday setup a IFTTT to send me a SMS if the RSS of a promotions website updates, something these travel sites could sell for a sub if they wanted to.
Personally when I'm looking for flights I'm not looking for travelling tips. I don't need photos nor do I need any editorial. And a map is an unnecessarily complex interface for what I do want to do.
I use Hipmunk for finding flights and I think it's a very functional product. Far superior as a product to what I would find on an airlines website.
I agree with the other comments made so far that the UI shown would be frustrating to use from a functional standpoint. I believe the "perfect" airline website would be something in between what is shown and what we have today. Something similar to Airbnb in terms of a nice UI that makes finding destinations and planning trips both functional and fun.
I travel a lot, and I can safely say that I don't like any of the current solutions much. It always takes me a long time to find a good pleasing flight. But their proposal doesn't really cut it, sadly.
Looks kind of annoying to use. I've usually searched out exactly what I want on a price comparison site and want the airline site to shut up and give it to me...
f-i, first and foremost, is an interactive agency, meaning they're all about selling their 'next' project; this leads me to believe that the site is probably a response to an RFP, or a really slick sales tool. As mentioned previously, this is a nice proof-of-concept for cutting-edge web UI work and interaction design (excusing obvious UX issues) that will '[sell] design consulting', but that's about it
I hope not, especially for visually impaired users, tablet or mobile users, or just for general ease of use. It's way easier for me to type in a location's name or airport code than try to figure out where it is on the map.
Here's an example: let's say I'm flying from Seattle-Tacoma (SEA) to Boston's Logan Airport (BOS). On a normal travel site, I type in SEA<tab>BOS<return>. Boom, done.
In this design, I'd somehow or another choose my origin, and then try to figure out which of the many major airports in the Northeast I'm actually looking for. My sense of geography of that area is pretty bad, so it would likely take me a while to figure it out.
Or let's consider the reverse: flying to the Pacific-Northwest. If you're zoomed out far enough, you're going to have a very small amount of distance between Vancouver, Seattle, and Portland. I guess you'll need to zoom in far enough to be able to visually differentiate them in order to drop the pin? OK, but why not just type in "Vancouver", "Seattle" or "Portland"?
This flight path map would be a nice infographic next to the traditional Leaving From/Going To web form, but it's not user friendly enough to be the only UI available. Most adults can't even locate their destination cities on a map.
Being a big fan of AirBNB, Hipmunk, and GoPexo, I think this concept is really exciting.
Unfortunately, I also think that it'll probably remain a concept for at least the next five years.
One of the biggest PITAs with travel is getting accurate and comprehensive flight and ticket information; with the exception of airlines running on Google's ITA, I believe most airlines are running on systems built with Fortran (and thus are difficult to interface with). While companies like Kayak are making this information more accessible, I would be really surprised if a real-world airline managed to build anything remotely like this any time soon.
As a "destination discovery" system, I think this would be ok. However, I'm not really the target market. I already have ideas of where I would want to travel to, I just don't have the time or the money. Seems that a site like this is designed for someone who has the time and the money but not the ideas.... First world problems.
Personally, if all airlines in the world would magically work with google flights, I'd be happy.
One area of booking flights that always annoys me is that I have to pick ONE airport to fly from.
I live near to 3 good viable airports (within a 2 hour drive). I would like to pick my preferred destination and select all 3 airports as start points, and then let the algorithm find the best prices from all 3 airports.
They often have a "+/- 3 days" option. They need a "+/- X hours drive time from LAT/LNG".
We see this sort of thing a lot. Creating something pretty doesn't necessarily give the user a nice experience. To be honest I've never found booking flights that hard, I normally just use a comparison site and it works fine. Designs like this seem a bit gimmicky and as book a flight normally involves spending quite a bit of money that's the last thing I want to see.
I thought this was real and was really really excited about it. I've been traveling for a pretty long time and booking an airplane has always been an annoying experience. Recently UI like Hipmunk and Adioso really changed a lot and I'm really grateful that hackers have decided to improve on flight research. If someone could do what f-i just did it would just be amazing.
In addition to all the other UI flaws this site has, I'm surprised nobody has mentioned that the site itself is about 15 "pages" long. I gave up scrolling after 30 seconds. I'm also very surprised that whoever designed the site thought that was a good idea-- I feel like that's a pretty obvious design no-no.
The proper way to appreciate this website is with Van Halen's "Right Now" in the background. Open this link in the background to see for yourself: http://grooveshark.com/s/Right+Now/3XdxpO
I'd rather that Airline Websites use their time in improving the usability of their site by using autocomplete suggestions instead of drop-down lists of hundreds of cities, save information in case of failed payments etc rather than make the site beautiful but non-usable.
Surprising how they made a nice UI but at the 1min mark they show eastern Asia with the shades of yellow reversed. The land mass should be darker than the ocean. It makes it look like the cities are on a land mass that is the shape of the ocean.
Blah blah blah, people know where they want to go before they are booking a damned flight, and Google Flights already lets you book them significantly more easily and cheaper than a single airline.
The map is highly inaccurate, I guess the designer used Illustrator (or similar vector tool) and applied some effects/transformations to the curves representing the borders.
The reason is that the domain details are so extensive and difficult that almost nobody can get past them. I like to describe both these problem domains as "fractal," because when you're 100,000 feet up it looks pretty simple, but the closer you get the more details there are. If I had a dollar for every hacker who told me how easy it would be to make a travel search web site or an email client, I'd be rich.
In both spaces, a very small number of players create the core technologies, and a much larger set of players layer on top of these cores. In travel, for example, you have the GDS companies (Sabre, Amadeus, etc.), ITA, and Expedia. Everybody else -- literally, everybody else -- layers on top of one these systems. Kayak? ITA customer. Hipmunk? ITA customer. Etc. This means the vast majority of players don't actually have to deal with the fractal domain complexity. And almost nobody has any clue there's any difference between say, an ITA and a Hipmunk, even though one has a million lines of code and runs thousands of machines and the other has a fairly standard website. (#)
Similarly, in email, a handful of players create real mail stacks. These are the usual suspects (Microsoft, Google, Apple, IBM) and a handful of others (Inky, Sparrow, Thunderbird) Everybody else -- literally everybody else -- layers on top of GMail or Outlook. Mailbox.app? Layer on top of GMail. Xobni? Layer on top of Outlook. Postbox? Fork of Thunderbird.
In the mail space, there are at least some decent open source libraries to use. In the travel search space there is literally nothing to start from but a blank sheet of paper. Carl de Marcken did most of the early work figuring it out for ITA, and it nearly killed him.
I actually seek out domains like this; either because they're defensible, or because I'm insane. I'm not sure which yet.
(#) If you're thinking it's necessary or even better to own the million lines of code and run thousands of machines, think again: Kayak created substantially more value for its shareholders than ITA did, with a lot less effort.