FlightCaster (YC S09) on AWS Blog(aws.typepad.com)
aws.typepad.com
FlightCaster (YC S09) on AWS Blog
http://aws.typepad.com/aws/2009/08/flightcaster-awspowered-flight-delay-prediction.html
6 comments
I don't get FlightCaster and I don't see how it can make enough money to justify external investment. It is not very useful to me to have a "maybe" about my flight - I will still go the airport on time and wait for the flight.
And furthermore, flights nowadays in my part of the world are on time. I've not been delayed in quite a long while.
FlightCaster has to be more than 95% accurate all the time, or its not useful. And I don't believe this tool is 95% accurate.
And furthermore, flights nowadays in my part of the world are on time. I've not been delayed in quite a long while.
FlightCaster has to be more than 95% accurate all the time, or its not useful. And I don't believe this tool is 95% accurate.
Hi Max,
Flightcaster may not be as useful for all travelers, but we believe it may be useful to a large enough population of travelers to be a real business.
Some business travelers can rebook flights in order to significantly increase the likelihood that they can make an important business meeting. In this case, it may be aimed at connecting flights, or at initial departing flights.
For others, having the information is useful, even if it doesn't mean rebooking. For example, I have traveled a lot for business and we joke about the value of the power outlets at the airports - another 20 minutes with power for my laptop is a big value in the airport. :-)
Lastly, we humans have to be careful about our bias toward making ambiguous numerical claims and extrapolating out from our limited experiences. The flights you have taken in your part of the world is a small sample size to generalize from. Depending on how you define us "delayed," somewhere between 20 - 30 % of U.S. domestic flights are delayed.
What numbers do you have to substantiate the claim that "flights nowadays in my part of the world are on time?"
Flightcaster may not be as useful for all travelers, but we believe it may be useful to a large enough population of travelers to be a real business.
Some business travelers can rebook flights in order to significantly increase the likelihood that they can make an important business meeting. In this case, it may be aimed at connecting flights, or at initial departing flights.
For others, having the information is useful, even if it doesn't mean rebooking. For example, I have traveled a lot for business and we joke about the value of the power outlets at the airports - another 20 minutes with power for my laptop is a big value in the airport. :-)
Lastly, we humans have to be careful about our bias toward making ambiguous numerical claims and extrapolating out from our limited experiences. The flights you have taken in your part of the world is a small sample size to generalize from. Depending on how you define us "delayed," somewhere between 20 - 30 % of U.S. domestic flights are delayed.
What numbers do you have to substantiate the claim that "flights nowadays in my part of the world are on time?"
I can understand that there is a category of travellers that would need something like this, but what I don't see is the money.
I.e, rebooking a flight is an emotionally taxing activity. It's like being in the supermarket and deciding if you should switch to the other line or stay in your line. It's a very small decision and the difference in time lost or gained is tiny, but emotionally, it's a pretty big decision. Rebooking a flight is also an emotionally expensive decision, and if I would ever trust flightcaster, it has to really be completely accurate all of the time. I would kick myself if I rebooked and something went wrong.
SO! We were talking about the money - the business is like a drill down list, and in each step, you have a smaller number:
Travellers > Travellers who care about being on time > Travellers who are internet saavy > Travellers who have heard of your app > Travellers who want your app > Travellers who will pay for your app.
With that depth in the list, you better be charging a pretty penny for this or, and this would be my personal preference, leave the consumer market and sell it to businesses as a B2B solution.
Selling to consumers is going to be futile - you may make money for 2-3 people to pay their bills, but you will not make money to justify an external investment.
I.e, rebooking a flight is an emotionally taxing activity. It's like being in the supermarket and deciding if you should switch to the other line or stay in your line. It's a very small decision and the difference in time lost or gained is tiny, but emotionally, it's a pretty big decision. Rebooking a flight is also an emotionally expensive decision, and if I would ever trust flightcaster, it has to really be completely accurate all of the time. I would kick myself if I rebooked and something went wrong.
SO! We were talking about the money - the business is like a drill down list, and in each step, you have a smaller number:
Travellers > Travellers who care about being on time > Travellers who are internet saavy > Travellers who have heard of your app > Travellers who want your app > Travellers who will pay for your app.
With that depth in the list, you better be charging a pretty penny for this or, and this would be my personal preference, leave the consumer market and sell it to businesses as a B2B solution.
Selling to consumers is going to be futile - you may make money for 2-3 people to pay their bills, but you will not make money to justify an external investment.
For heavy business travelers, the question is not whether to go to the airport or not, but whether to rebook or not, and they decide that based on probabilities.
Flying is an uncertain business time-wise, and I think most business travellers have factored the uncertainty into their schedule, and are not constantly looking to see if they should rebook or not.
The technology is good, and it's a useful tool. But technology does not give the money, it's the customers who do.
The technology is good, and it's a useful tool. But technology does not give the money, it's the customers who do.
I travel quite a lot. I don't think flightcaster would be much help for domestic itineraries, from JFK to SFO, for example.
It doesn't deliver much value for the work week frequent flyer (those who travel for 5 work days). Critical stuff happens on Tues/Wed/Thurs for just this reason. It's no big deal if I get in late on Monday--I just sit in the lounge and hack or answer email.
It doesn't deliver much value for the work week frequent flyer (those who travel for 5 work days). Critical stuff happens on Tues/Wed/Thurs for just this reason. It's no big deal if I get in late on Monday--I just sit in the lounge and hack or answer email.
When you have families every hour counts. It does matter if I get in late on Monday, even though I am perfectly find hacking away in the lounge
I travel quite a bit as well. I've gotten pretty savvy at avoiding airports that are busy. I can see the real value for this application to business users, however like you, I'm not convinced this will be a huge market. But that's maybe because I haven't calculated the hard numbers (how many people are actually afflicted by this problem and are willing to pay money to solve it).
That said, I'm convinced there is an enormous demand for prediction services. If someone could have predicted Boeing's extreme manufacturing delays with the 787 looking at supply-chain optimization, that alone would have saved billions.
That said, I'm convinced there is an enormous demand for prediction services. If someone could have predicted Boeing's extreme manufacturing delays with the 787 looking at supply-chain optimization, that alone would have saved billions.
If you know your flight has a certain probability of being late, are you going to risk arriving late at the airport and missing the plane if and when FlightCaster gets it wrong?
Another inherent problem is related to the fact that until the airline admits that the flight is going to be late (and sometimes even after that point) they still require you to check in on-time or risk losing your seat and pay extra for rescheduling. Then, what good is the information?
As customers of the airline industry we have been trained to adjust our behavior and expectations regarding timing without being able to influence the industry. The service will NOT and cannnot change it.
Given the factors involved, it is unlikely that the model used is able to explain substantial amount of the variance. Hence, the ability to predict low probabilities.
Sorry, I will not invest a penny in a company like this.
Another inherent problem is related to the fact that until the airline admits that the flight is going to be late (and sometimes even after that point) they still require you to check in on-time or risk losing your seat and pay extra for rescheduling. Then, what good is the information?
As customers of the airline industry we have been trained to adjust our behavior and expectations regarding timing without being able to influence the industry. The service will NOT and cannnot change it.
Given the factors involved, it is unlikely that the model used is able to explain substantial amount of the variance. Hence, the ability to predict low probabilities.
Sorry, I will not invest a penny in a company like this.
The transitions I made from FORTRAN to PL/I to C to C++ to Java to Perl to PHP were each pretty painless
How do you go from Perl to PHP without pain? You lose pretty much every good language feature in the process, and significantly reduce the number of tools and libraries you have access to. This remark blows my mind.
How do you go from Perl to PHP without pain? You lose pretty much every good language feature in the process, and significantly reduce the number of tools and libraries you have access to. This remark blows my mind.
What I meant was that much of what I learned with language N was applicable to language N+1.
Some of these new programming models don't seem as if they build on concepts that are familiar to me.
Maybe I'm overstating this, but I get the sense that we are approaching some sort of discontinuity where the developers who can take these new, non-procedural, declarative, loosely typed, implicitly parallel systems and make them dance and sing will be light years ahead of the last generation and their pithy procedural languages.
Some of these new programming models don't seem as if they build on concepts that are familiar to me.
Maybe I'm overstating this, but I get the sense that we are approaching some sort of discontinuity where the developers who can take these new, non-procedural, declarative, loosely typed, implicitly parallel systems and make them dance and sing will be light years ahead of the last generation and their pithy procedural languages.
Is pithy really the word you mean to use?
Not sure, but I've been up for 20 hours and can't think of anything more to say. Goodnight!
The transitions I made from FORTRAN to PL/I to C to C++ to Java to Perl to PHP were each pretty painless
That just means he's still writing FORTRAN in PHP syntax :P
That just means he's still writing FORTRAN in PHP syntax :P
To support the problem identified earlier: "75% of all significant delays (those over an hour) are first posted by the airlines within 30 minutes of departure." Yet, in most cases you have to check in at least an hour before departure.
So, this statics actually works against your service.
The point here is that we tell you 6 hours before departure--long before the airline posts anything and long before you check-in.
The stats FlightCaster collects is like the stats on road accidents: if you go by route A on Ford you have a 90% probability to broke the car (and be late), so you better switch the route or the car.
How many will do that?
How many will do that?
25% of all flights are delayed. If you're on the last flight of the day--you have a 36% chance of delay.
That's 400,000 people per day in the U.S. alone that are delayed without much or any advanced notification. 75% of all significant delays (those over an hour) are first posted by the airlines within 30 minutes of departure.
For those that don't want to risk missing a meeting or getting home for dinner, advanced notification means you can change flights or at least warn your family and colleagues.
Stay tuned--we'll soon be posting all our accuracy metrics at http://blog.flightcaster.com.
For now, read more about how delays occur and why airlines report them so late at http://flightcaster.com/faq