Disclaimer: This is a throwaway account. I'm a person who has had a travel startup. I've decided to halt, once Google has finally acquired ITA Software.
Travel is one of the world's largest industries (~5th), and online sales are its significant part (~10-30%, depending on the market), growing strongly. Online travel agencies (OTA) are among the few companies on web that get real money (~$150 per sale) from customers (it is relatively easy to make serious revenue).
Online travel sales consist mostly of flight bookings, and hotel reservations.
On these markets, there are roughly three categories of players. Airlines and hotels _provide_ the inventory, that is flights and hotel rooms. Computer reservations systems (GDS) _manage_ the inventory. Online travel agencies _sell_ the inventory.
Specifically, there are thousands of airlines and hotels (e.g. Hyatt, Lufthansa). However, there are only three major GDS operators in the US and Europe (i.e. Amadeus, Sabre and Travelport), as well as only few big OTAs (i.e. Expedia, Travelocity, Orbitz, Priceline, Hotels.com, and Opodo).
Few decades ago, before Internet, airlines and hotels were unable to sell inventory on their own. It definitely takes more time to set up an office than a connection between two airports. Thus, it made sense to use travel agencies for this purpose.
Over time, airlines and hotels became also unable to manage their inventory. Synchronization of reservations between thousands of third parties is a non-trivial task, not a core competency of involved companies. Thus, it made sense to use a middle man.
That's what GDS systems do. They manage inventory, what includes reservations (PNR), its availability, prices, and exchanging data with others. As far as I know, right now, airlines and hotels pay them for the service, and for each reservation made (~few bucks).
Internet has complicated things a bit. Travel agencies are no longer so vital, as both airlines and hotels are able (at least, they think so) to sell the inventory on their own. Unfortunately, they were too slow again, and OTAs has emerged meanwhile (Expedia), giving the second youth to GDS companies.
The ecosystem is like an old marriage couple, although a threesome. Each party hat hates each other, but there's no other way around. To oversimplify a bit, OTAs have _customers_ (traffic), GDS systems _manage_ the inventory (with airlines, hotels, and OTAs) and both airlines and hotels _provide_ the inventory, after all.
From time to time there's an affair. Low-cost airlines try to distrupt the market by selling tickets so cheaply partially because they sell them directly. Major OTAs, like Expedia, partially grow to a GDS category. Some airlines or hotel chains withdraw from GDS systems, and return eventually with negiotiated better fees. However, it's mostly business as usual.
Today, if you want to start an online travel agency you have to speak with a GDS company. Depending on your market, it might be Amadeus, Sabre or Travelport. After a long selling process, you get access to the system, and you can start selling the reservations.
What's important, though, nearly all systems used today were created a decade, or two, ago. As core competency of GDS companies is in selling, then, as far as I know, they outsource the software evelopment to third parties, and it's not that easy to innovate on a critical part of the world's infrastructure.
What you end up with, then, is an access to an undocumented API that lets you to search, and manage your reservations. Insiders are used to the quirks, like waiting few seconds until you get the response, random issues, or hinting the system so you get a better response than others. Importantly, you're actively discoured to cache the data, as the prices change rather frequently.
The critical part here is search. It's a mathematically non-trivial problem to very quickly find rates within thousands of connections, definitely beyond technical know-how of GDS operators, airlines and hotels. ITA Software has managed to get access to the inventory and while, as far as I know, they do not sell resevations, they've created a much better (faster) search.
Meanwhile, few years ago, metasearches (e.g. Kayak, Hipmunk) emerged. Smart folks have realized that the competition is on price, customers look for a single place to compare prices, and operate under assumption that what really counts is traffic. From both customers and metasearch perspective, it does not make that big difference where do they buy the reservation from, an OTA or directly from an airline.
So, here we are today.
As a beggining travel agency, you likely have to pay annually for access, and for each request made, especially if you exceed the quota negotiated with the GDS company. Few years ago you were able to make profit by incurring a transaction fee to each ticket sold, but now transaction fees are nearly non-existent, and it's more frequent to rely on provisions from GDS companies, and, sometimes, airlines.
What's your competetive advantage? Basically, you cannot provide much better product than your competitors, as everyone relies on the similar legacy GDS system that returns the flight details rather slowly. Most of the time, only choices are either to show the results a bit differently, or bet on more trustful brand.
The focus is on efficiency. Profit per ticket is so slim, so cost of customers acquisitions is what really matters. OTAs, metasearches and, increasingly, airlines, together with hotels, master SEM, SEO, and other forms of advertisement (newsletters, banners). They live and die by the numbers. If you've figured out how to scalably make $1 more profit on each reservation made, you're covered for some time. The novel methods are, obviously, eventually realized by others, too.
In this race to the bottom there's one clear winner. Google's AdWords is a major source of traffic for all parties, and I bet that they already make the biggest profit off each reservation made. Once Google has acquired ITA Software, they now have both traffic and the inventory.