4 ms·

Apply HN: 128Highstreet – A Trip Planning App

2 points by Sreyanth 11y ago

Problems: 1. Planning a trip to a new place is usually tiresome, with reading a lot of reviews on TripAdvisor, asking a lot of relatives or friends for suggestions, using the existing trip planner websites that suggest pre-generated itineraries - that probably worked for someone else. 2. After carefully considering a lot of choices, the user finally decides on a trip plan. But things may change - there might be new events coming up in the travel destination which the user doesn't want to miss or there might be last minute issues like a traffic jam or a bad hotel experience. The user is forced to stick to the previously decided plan and compromise! What we do: 1. We do all the painstaking work of understanding the reviews on TripAdvisor and similar sites, learn what the user's network is speaking about various travel destinations, the ones the user like the most - take into consideration the user's budget, seat preferences, meal stops, gas stations, driving time, sleeping habits, previous travel history, nationality and visa requirements etc., and build the most suitable & easily editable trip plan for the user. (The final aim of this app is to make sure the user won't press the edit button. i.e., the itinerary generated is perfect for the user). We track the user's trip and suggest Plan B, just in case 2. We want to offer the user with hotel/car/event/flight bookings with no booking or cancellation charges. Is this even feasible? 1. Yes. We are in a process of building an AI engine that makes the trip planning possible. The aim is to get the user choose a perfect trip plan without spending hours to input all the details. 128HS aims to capture the user's family members'details as well and keep learning when to show what! 2. The same AI engine calculates the probability of a user cancelling a hotel, a flight or any event, and tries it best to minimize such a probability. Minimizing this probability would mean ensuring the quality from the partners, using historic data and the likability by the user. Sreyanth http://128highstreet.com

d--b 11y ago
Sorry, I don't want to be a pain, but you should read this: http://blog.garrytan.com/travel-planning-software-the-most-common-bad http://blog.garrytan.com/travel-planning-software-the-most-c... And sorry for the negativity again, but AI as it is today is not capable of understanding what people want. I mean, even real people are not able to understand what other people want or like. The tradeoff between quality/convenience/price is extremely subjective. People are complicated. That said, I wish you luck with the project, and I'm hoping to be wrong.
- Sreyanth 11y ago
  We do agree people are complicated, and it is extremely important to get it right the first time, especially when the user's mood is bad. But, we are looking at trip planning as a context aware recommendation problem that gives out ranked results which match the user's preferences. We hope to make our trips as relevant to people as Google search results are.
aacook 11y ago
A while back I was thinking about a trip planning product and my enthusiasm was deflated when I read Gary's post. I remember one of my main take-aways was that travel just isn't something people do very often. So even if they use and love your product, the chances of them remembering you the next time they need to plan a trip (a year later?) are slim. I ran into this same problem working on a product rental marketplace (bouncy castle rentals, wedding tents, etc) and it was definitely a struggle. We had to constantly acquire new users. That said, I wonder if there's something here if you focus on business travelers. (1) Some business people travel quite often and (2) Lots of people pay for travel agents or use internal resources to plan their trips. Recently Paul English, the co-founder of Kayak, launched Lola. It's sort of like Magic or Operator but travel-specific. May be worth checking out: http://lola.co http://lola.co. Another area I like about travel is safety. When you're traveling to a foreign place, it's easy to assume a given area is safe ("It's only 2 miles from the city center!"). I'd pay money to have access to a real local to warn me of areas to stay away from and let me know of things I shouldn't miss.
- Sreyanth 11y ago
  We want to retain our customers by providing them the best "benefit - cost" value when compared to the competition. To make sure people remember us the next time they plan a trip, we surely want to stay in touch with the user by suggesting new places, trip plans etc., (to be noted that we get this information from the user's email and calendar - we would be using push notifications to contact the user. Some website suggests, forgot which one, that most people go on a trip when suggested via email or text). And we hope that our choice of the name 128Highstreet will make it appear among the top in the apps menu in many mobiles, thus catching the user's attention every time they open up the menu. We were pretty much looking at various types of customers who would be needing a product like 128HS - 1. Consultants (who got their travel scheduled, and have some leisure time as per their schedule) 2. Casual travelers (like the normal customer, who cares to travel once or less) 3. Students (grad trips, weekend trips) 4. Tour operators (people who sell packages to customers) 5. Travel agents (who earn commission on the amount of money the end customer ends up spending) We are in a way similar to Lola, but we want to eliminate humans from the process. No doubt, we would be needing human support in the initial days to get going. At 128HS, we care about the customer's experience - and that's exactly how we are planning to not let the customer leave us. We plan to include safety and must visit places in the itinerary - we also plan to warn the user if an unsafe place is planned to be visited at an unsafe time. To speak more, since we track the user's trip (using the GPS on phone, 128HS is a mobile app BTW), we will be able to, in real-time, suggest changes to the user and cancel/book new hotels/events/flights without any booking/cancellation charges.
brudgers 11y ago
How does 128highstreet collect information about the user and their family in order to provide individualized recommendations?
- Sreyanth 11y ago
  We want to analyze the user's connections and emails to figure this out. We do understand that there would be a lot of privacy concerns in this regard, and hence won't like to save any such private data on our servers. Once we derive this information, we present it to the user to confirm & save. If a user A adds user B (who already uses 128HS), then we need not worry too much about learning about B as B is already a 128HS user. The problem is with those family members who don't use 128HS. We also encourage the users to share interesting plans/places/events with their contacts (which the user thinks their contacts might like), we keep learning more about their contacts. We do realize that the same user behaves differently / has different preferences when on a grad trip vs a family trip. We want to generate the network graph accordingly. On another thought, we were also looking at gamifying the entire trip planning exercise among family members with incentives for inviting family members to plan trips collaboratively - just an idea, didn't think through too much in this direction.
buss 11y ago
> We do all the painstaking work of understanding the reviews on TripAdvisor and similar sites How will you scale this? How will you customize results per user? What experience do you have building AIs?
- Sreyanth 11y ago
  > How will you scale this? We are looking at building a classifier that classifies reviews collected through TripAdvisor's content API and Google's search API. We want to cross-match the classification results, against the ratings (very crude, but a good way to start I guess). > How will you customize results per user? We would build packages that meet the user's preferences (which are subjective - we would learn these over time due to the cold start problem) and rank them appropriately, based on a large set of parameters. This, I suppose, would be a very nascent version of Amazon's A9 or Google search ranking. We will try our best to keep learning and improving based on the users' interactions. > What experience do you have building AIs? I haven't developed any in-production AI. I wrote a few AI programs for my AI course. Wrote a few classifiers to participate in big data competitions. I also wrote one such AI program to build bundled packages as part of experiments for my paper (https://www.computer.org/csdl/proceedings/icdmw/2012/4925/00/4925a843.pdf https://www.computer.org/csdl/proceedings/icdmw/2012/4925/00...).

2 points by Sreyanth 11y ago

d--b 11y ago

Sorry, I don't want to be a pain, but you should read this: http://blog.garrytan.com/travel-planning-software-the-most-common-bad http://blog.garrytan.com/travel-planning-software-the-most-c... And sorry for the negativity again, but AI as it is today is not capable of understanding what people want. I mean, even real people are not able to understand what other people want or like. The tradeoff between quality/convenience/price is extremely subjective. People are complicated. That said, I wish you luck with the project, and I'm hoping to be wrong.

Sreyanth 11y ago
We do agree people are complicated, and it is extremely important to get it right the first time, especially when the user's mood is bad. But, we are looking at trip planning as a context aware recommendation problem that gives out ranked results which match the user's preferences. We hope to make our trips as relevant to people as Google search results are.

aacook 11y ago

A while back I was thinking about a trip planning product and my enthusiasm was deflated when I read Gary's post. I remember one of my main take-aways was that travel just isn't something people do very often. So even if they use and love your product, the chances of them remembering you the next time they need to plan a trip (a year later?) are slim. I ran into this same problem working on a product rental marketplace (bouncy castle rentals, wedding tents, etc) and it was definitely a struggle. We had to constantly acquire new users. That said, I wonder if there's something here if you focus on business travelers. (1) Some business people travel quite often and (2) Lots of people pay for travel agents or use internal resources to plan their trips. Recently Paul English, the co-founder of Kayak, launched Lola. It's sort of like Magic or Operator but travel-specific. May be worth checking out: http://lola.co http://lola.co. Another area I like about travel is safety. When you're traveling to a foreign place, it's easy to assume a given area is safe ("It's only 2 miles from the city center!"). I'd pay money to have access to a real local to warn me of areas to stay away from and let me know of things I shouldn't miss.

Sreyanth 11y ago
We want to retain our customers by providing them the best "benefit - cost" value when compared to the competition. To make sure people remember us the next time they plan a trip, we surely want to stay in touch with the user by suggesting new places, trip plans etc., (to be noted that we get this information from the user's email and calendar - we would be using push notifications to contact the user. Some website suggests, forgot which one, that most people go on a trip when suggested via email or text). And we hope that our choice of the name 128Highstreet will make it appear among the top in the apps menu in many mobiles, thus catching the user's attention every time they open up the menu. We were pretty much looking at various types of customers who would be needing a product like 128HS - 1. Consultants (who got their travel scheduled, and have some leisure time as per their schedule) 2. Casual travelers (like the normal customer, who cares to travel once or less) 3. Students (grad trips, weekend trips) 4. Tour operators (people who sell packages to customers) 5. Travel agents (who earn commission on the amount of money the end customer ends up spending) We are in a way similar to Lola, but we want to eliminate humans from the process. No doubt, we would be needing human support in the initial days to get going. At 128HS, we care about the customer's experience - and that's exactly how we are planning to not let the customer leave us. We plan to include safety and must visit places in the itinerary - we also plan to warn the user if an unsafe place is planned to be visited at an unsafe time. To speak more, since we track the user's trip (using the GPS on phone, 128HS is a mobile app BTW), we will be able to, in real-time, suggest changes to the user and cancel/book new hotels/events/flights without any booking/cancellation charges.

brudgers 11y ago

How does 128highstreet collect information about the user and their family in order to provide individualized recommendations?

Sreyanth 11y ago
We want to analyze the user's connections and emails to figure this out. We do understand that there would be a lot of privacy concerns in this regard, and hence won't like to save any such private data on our servers. Once we derive this information, we present it to the user to confirm & save. If a user A adds user B (who already uses 128HS), then we need not worry too much about learning about B as B is already a 128HS user. The problem is with those family members who don't use 128HS. We also encourage the users to share interesting plans/places/events with their contacts (which the user thinks their contacts might like), we keep learning more about their contacts. We do realize that the same user behaves differently / has different preferences when on a grad trip vs a family trip. We want to generate the network graph accordingly. On another thought, we were also looking at gamifying the entire trip planning exercise among family members with incentives for inviting family members to plan trips collaboratively - just an idea, didn't think through too much in this direction.

buss 11y ago

> We do all the painstaking work of understanding the reviews on TripAdvisor and similar sites How will you scale this? How will you customize results per user? What experience do you have building AIs?

Sreyanth 11y ago
> How will you scale this? We are looking at building a classifier that classifies reviews collected through TripAdvisor's content API and Google's search API. We want to cross-match the classification results, against the ratings (very crude, but a good way to start I guess). > How will you customize results per user? We would build packages that meet the user's preferences (which are subjective - we would learn these over time due to the cold start problem) and rank them appropriately, based on a large set of parameters. This, I suppose, would be a very nascent version of Amazon's A9 or Google search ranking. We will try our best to keep learning and improving based on the users' interactions. > What experience do you have building AIs? I haven't developed any in-production AI. I wrote a few AI programs for my AI course. Wrote a few classifiers to participate in big data competitions. I also wrote one such AI program to build bundled packages as part of experiments for my paper (https://www.computer.org/csdl/proceedings/icdmw/2012/4925/00/4925a843.pdf https://www.computer.org/csdl/proceedings/icdmw/2012/4925/00...).