Posted in: Uncategorized

Thoughts on Social Travel – My Own Social Graph Isn’t Good Enough

I have been doing a lot of travel lately, some for work and some for fun. The one thing that continues to stun me is the dominance of TripAdvisor as a source of reviews whenever I go to a new place and start looking for restaurant or hotel reviews. It reminds me a lot of looking for restaurant reviews and only finding Citysearch before Yelp came along and shook things up. While generic search queries might return a diverse set of sources, specific queries (restaurant x in town y) almost always have several TripAdvisor responses in the top set of search results.

I know Yelp is continuing to expand globally, and it couldn’t happen fast enough for my tastes. When I saw TripAdvisor release their Instant Personalization, I thought it might be cool. But it doesn’t solve my problem. Some time you need to see someone else’s solution to really figure out what it is that you’re looking for. I’ve heard some of my friends who travel say that they don’t trust TripAdvisor rankings because they might be written by employees of the places being reviewed, by people who’ve never been to the place in question, or generally don’t feel “real” (whatever that means). I don’t care so much about the trustworthiness of the reviews and reviewers – I care more about relevance and utility. If someone astroturfs a review but it turns out that it’s a place I like, that’s not as bad as a genuine review whose conclusions strike me as odd.

Whenever I’m planning a trip to a new place, I always ask myself the same simple question:

For someone with my tastes and interests, where should I eat and where should I stay when I’m in a new town?

There are a number of reasons why I think the TripAdvisor status quo is ripe for some disruption. I’ll try to summarize them below:

1. My perception of TripAdvisor’s search results is that they are self-reinforcing
On my last trip, I sat next to a couple who described TripAdvisor as “their bible” – they only go to places that are highly rated. So I went to a number of the highest-rated places and they were largely filled with out of town people, many of whom were either clutching guidebooks, printouts of TripAdvisor reviews, or made comments that led me to believe that their decision to show up was influenced by something like TripAdvisor (if not the service itself). I’m sure some of those people will go back home and dutifully review those places. Which will add more reviews to the top places. And thus the cycle continues. I don’t know much about the TripAdvisor ranking algorithm or how it works, but I think most people tend to look for the highest-rated places and start there. If for some reason those places don’t seem interesting or appeal to them, they move on.

There’s nothing wrong with this approach – taking into account the density of reviews and the ratings seems like a rational way to do things. But it does have the overall effect of driving many casual visitors to the same set of places and concentrating reviews in a set of locations that already have lots of reviews.

One small hack I’ve been using to get around this is to just skip most everything on page 1. Those places are probably good, but there are so many other places that could be great, will probably have fewer tourists, and might be more my style. I usually start hunting on page 3 of the results, where the review density is generally lower and then I ask the concierge or some local people if the places that look interesting to me are any good.

One of the things I’ve always really liked about Yelp reviews is that I can generally get a lot of context about the reviewer. Is the person from San Francisco? Have they reviewed other places I like? Do they reference other restaurants in their reviews? Those things are what help me figure out the reviewer’s context, bias, etc for reviews.

2. I can’t make heads or tails of most TripAdvisor reviews once a place has been reviewed about 20-30 times
If you read the reviews of the top places in any given city, you often end up with the same pattern:
1. A bunch of really positive reviews about the place
2. A few people who had wildly different experiences and call the place overrated, terrible, or not worth a visit
3. A bunch of reviews in between that are very measured in tone

That’s simply not useful to me. Yes, you could read through 10-15 reviews to try to get a complete picture of the restaurant or place in question. That’s time consuming. And at the end of the day, you don’t know anything about the people behind those reviews. You don’t know where they like to stay, what their standards are, what they’re accustomed to in terms of service, and what their idiosyncracies are. I bet there’s probably a review buried in there written by someone like you – I challenge you to find it.

There’s another, more subtle problem that I often find with reviews. Many times, people make comments about the quality and price of the food, especially at restaurants. I think it’s fair to say that not everyone has the same idea of what constitues “cheap” or “expensive” or “best meal we had in town” – this can make it extremely difficult to figure out what to expect. Sometimes people will give you clues in their reviews – they’ll say things that tip their hand about what they consider to be fair prices or other places they’ve been that rank highly to them. This drives me nuts. Even services that include some kind of rough price expectation, the actual bill can vary wildly depending on what you order. I have another hack for this. I often tend to look for reviews written by people who travel without children and as a couple, which is a rough proxy for how I often travel. But that takes some detective work on my end to figure that out. It should be easier.

3. Using my social graph is not good enough – I want people with my tastes, not people I know.
While I do put a lot of stock in personal recommendations and referrals when I’m traveling, there’s no guarantee that my friends will have the same travel or lodging tastes that I will. Some are into ultra-deluxe. Others are really into hostels and low-budget accommodations. Ditto on food. Like most people, I’m generally able to filter the appropriateness of a recommendation based on the source.

But that doesn’t work at web scale. There are probably many other people out there with my tastes in dining and lodging who are not inside my social graph. Those are the people who’s opinions and thoughts I’d like to read. As I mentioned above, one of the hardest things to figure out about TripAdvisor is who the person is behind the review. I don’t mean whether they’re a shill or a real person – I mean you don’t know what they consider good, bad, or other.

I haven’t used Hunch, but maybe that’s something that Hunch or a similar service can help deliver.

3. There’s lots of other really good sources of data out there that could be used to help me find similar people.

The reason I’m so excited about this problem is that we now have more data sources that could be used to address this problem. For example, I have almost 2,000 foursquare checkins and a decent number of Facebook checkins too. That tells you a lot about the places I like to go and the intensity with which I like to visit them. Why not use that data (as the corpus grows) as a key input for restaurant or travel recommendations? Look, you know where I like to go and you can find other people who like to go to those places too. It seems like a critical, and previously-unavailable, data source that could be an important foundation for a next-generation service here. I have to believe this is on the roadmaps for foursquare and Facebook

Okay, that’s a lot of stuff to post. Hope you enjoyed it. As always, feel free to leave some comments.