HN user

almacmillan

15 karma
Posts0
Comments12
View on HN
No posts found.

I think the issue is that ITA, Skyscanner et al who built their own tech freed themselves from slow old GDS platforms (which were ancient mainframes and designed for travel agents). They consume the schedules, the fare rules and then they will use 'direct connect' XML or JSON feeds directly from the airlines (if they are available, not all airlines do them), then a GDS, and sometimes they will have to scrape a website where a LCC doesn't use GDS or give a metasearch site api access to a) regularly check fares for a given flight, time & date and b) to do a 'live' search e.g on skyscanner, it will often say 'price was checked 30 mins ago' (which is the cached and stored in the db price) and then when you select it and a modal popups up it does an ajax call to check if that price is still live. But to more simply answer your question - many airlines love 'direct connect' as it bypassed the GDS who, remember, they pay fees to and have had them over a barrel for decades. And metasearch sites prefer it to. It's a direct relationship. It is a really complicated industry and I often read gross over-simplifications of x site versus y site.

Rubbish! ITA's matrix site is probably the best air search site there is and is just a demo of their api. Yes ITA is expensive but you are paying for search, they don't do booking. But the search is awesome because unlike the collective GDS, they are purely a search company. Whereas GDS are reservation systems and they make their money via fees paid to the GDS by the airlines. Search as an api business is a secondary concern/revenue stream to them.

And why are they inherently unprofitable?Sites like Kayak, Skyscanner, Adioso are meta search sites only that make money by sending the user to another site to make a booking. They then make a small affiliate commission if the user then books their flight. Airlines actually really like good meta search sites because they acquire customers cheaply and avoid GDS fees altogether so are more likely to pay commission including your example of Qatar Airlines who pay 2% per booked flight to affiliates see http://www.qatarairways.com/ae/en/affiliates.page and not ZERO as you say. So they have zero need for the GDS unless an airline has no direct connect api in which case the GDS is that api source usually for availability (if you have implemented the schedules and fares locally) if not then a dumb search to the GDS is possible but a crap, slow, expensive way to include an airline in your site and no use to Adioso who need fast local graphed, cached data for expansive NLP searches.

Adioso began as a single source for price search to low cost carriers and the low cost carriers are not in the GDS because they won't pay the prices.So I presume they scrape (spider at regular intervals) the low cost carrier prices from the airline websites but possibly they have api direct links. Tom could tell us. (Tom I really want an in depth post on wingtip!) But to do fast real time search for hundreds of airlines you need more than a fast schedule (time dependent graph) you also need the complex fare rules worked out and programmed. All of that data is available but costs. The last bit is you then need to check availability via wherever you can book the ticket (GDS or direct to airline) and cache it for speed. I think that the fact that Tom links to Vayant's price cacheing api it is an admission that it isn't raw data that's lacking it's a higher level api where fares and availability are already worked out. Because seriously to do that yourself is a huge time & engineering challenge. It took ITA years. Adioso have the fast Graph and can get the schedule data. Vayant is offering an off the shelf solution to do fares and availability. Could work.

Ah a GDS romantic. GDS systems suck at search. Tell me? Why can't you even do these type of searches on the Amadeus flight search page? http://www.amadeus.net Maybe you would be kind enough to do a search from somewhere to 'anywhere' or 'Europe' from 'any date' to 'any date' on sabre and post the video. Bet you can't. Great flight search only happened when ITA liberated search from the GDS dinosaurs.

Nah, in modern flight search, ITA & Vayant take all the raw flight data like schedules, fare rules etc which they buy from companies like OAG and Innovata & IATA. GDS increasingly are nothing more than a source for cacheing prices when a 'direct connect' connection to the airline is not available or where the end consumer (an ota or travel agent for example) will be purchasing via GDS. Certainly no GDS has tech that does open ended real time Graphed queries like Google Flights, Adioso's Wingtip, Vayant, Skyscanner can do

Thanks for this series Tom, you guys are trying to succeed in a really challenging technical sector (Travel) and I, for one, really appreciate your honesty. I think that some of the confusion & criticisms I see in other comments are that after reading your pieces we all rush off to try adioso and some observations predominate:

# The ui is quirky on different screen sizes # The calls to Expedia are SLOW # The natural language search is perhaps only one search interface and not necessarily the best # The core bit you guys have worked on 'wingtip' is sort of lost in this version of Adioso because of trying to increase coverage # You could have utilised existing api's like ITA's QPX, Everbread's Haystack, Amadeus, even Skyscanner's api and increased coverage whilst being much faster. # The example link at the end that presumably is the answer to the question "For the first time, we’re seeing signs that the flight data we need is becoming available, in the right format" is totally misleading. Vayant are an ITA competitor not a raw data provider. They have managed to achieve in their price cacheing tech what you have only achieved for lots of low cost carriers. # And this last point is my biggest confusion about Adioso and what I think is your internal confusion. Are you a tech platform for flight data or a consumer facing flight website. Because Kayak, Hipmunk (and previously Orbitz erc) are B2C consumer sites that get data from ITA, Amadeus etc. Vayant, ITA etc are B2B data platforms - tech companies. Only skyscanner truly is both I think. They are a B2C site that last year enabled the selling of over £2 Billion worth of tickets but which is powered by its own 'Graphite' Graph and price cacheing databases.

The embarrassment I feel for you is when you say that you are making a travel site 'even when it's impossible'. I cringe for you! You haven't built a site even as good as Skyscanner's and you're saying pretentious statements like that? I know startups are hard and this is for a book so poetic license is to be expected.

I say ditch the BS and get back to reality. You could attach your natural language interface to someone else's api (like Vayant's) and have a better B2C product or keep improving 'wingtip' but this all seems like a confused desperate fudge.

Incidentally, when are you going to talk in detail about 'Wingtip' it's the bit that most interests me. It IS Adioso. I really do wish you guys every success though. What you are doing is needed but ditch the slow ass Expedia api.

There are 61A videos on ITunes U but it's hard to tell if they are from the Python version of the course? Anyone know if there are videos of the lectures kicking about from past years of the Python version?

We http://www.atomised.coop are a worker's co-operative too and a startup. There are ways to raise finance in the UK certainly because co-ops can have almost any legal structure. e.g we are a limited company with the ability to issue Class B shares (the class A ones are the worker/owner equity). You also get multi-stakeholder co-ops where equity can be divided between investors, workers, customers even. It's called the 'somerset rules' or more commonly multi-stakeholder co-operatives. See http://somerset.coop/somersetrules A firsthand example I know is a train company startup. They divided equity between investors 25%, workers 50% and consumers 25%. I've always thought co-ops could really marry well with open-source & startups but startups know nothing about coops and coops know little about startups.