Make the Internet Smarter at Helping Us

Posted on Feb 1, 2011 | 96 comments


Recommendation Slicing

The Internet has a problem.  It’s one you know well. The signal-to-noise ratio is getting out of hand.  In non-geek speak this means that there is so much information out there now it’s hard to separate the good stuff from the rest.

We all now visit UGC sites to learn about products & services before we use them. I’ve been thinking a lot about this over the past two years and bending the ears of any entrepreneur who will humor me to hear what I think the solution needs to look like.  I call it “recommendation slicing.”  I guess I sort of want “Pandora for everything else.” Why can’t I have it?

Here’s the problem and the solution set.

Problem
So I go to Trip Advisor and I want a recommendation to stay near Laguna Beach. I was thinking about the St. Regis so I went to see what people thought. Years ago we had been to the Ritz Carlton nearby and we were saved by Trip Advisor because a commenter had mentioned that one of the wings was under renovation so we were able to book the right side of the hotel. So I’m drawn to UGC reviews. But, oy, is it serious effort.

So there it is – the St. Regis, the number 28 recommendation in Orange County.  Hmmmmph. I thought it would be nicer?

And then I notice in the list of hotels above it: Best Western, Doubletree, Residence Inn, Howard Johnsons … all score higher!  Now, I’m not a hotel snob (ok, I kind of am) because I wasn’t born with a silver spoon in my mouth, but at 42 when I want to take a nice vacation I’m looking for a nice vacation.

And it dawned on me – every freaking recommendation website out there is just one big amorphous mass of group recommendations that don’t relate to me. They are the wisdom-of-the-crowd, sure, but a different crowd than I hang out with these days.

So I had to find out exactly what WAS wrong with the St. Regis. I started reading the comments of reviewers. It seems that people were complaining that there was a kiddie pool too close to some of the rooms and there were people with their kids yelling and screaming at 8am.  “Never stay here again!”  This was a few years ago when I was in the kiddie-pool phase (now we’ve luckily graduated to boogie boarding!).

All I was thinking was, “Hallelujah! Yippie! My kids can yell at 8am!  Sign me up, baby!”

And of course I know that some 25 year-olds are thinking that they’d rather stay in somewhere more hip, more alive at night & probably a bit cheaper.  That was once my demographic, too. We’re the same in some ways (my wife says my real age is still 13) but we’re different in others.

Food is no different. Whenever I’m trying out new restaurants I always feel compelled to check Zagat & Yelp and then STILL email friends to ask their POV on my top 3-4 choices. Cobbling together the historic wisdom (Zagat) with the edgier wisdom (Yelp) with my friends input gives me a pretty good triangulation and seldom fails me. If I listed to Yelp alone I’d mostly eat at the 25-year-old hangouts all the time.

Neither Zagat nor Yelp seem to do a good job of capturing what the new trendy, up-and-coming restaurants are, which is important for people who live in big urban cities where restaurants change like fashion styles. Luckily I have access to Kelly Amoroso in our office who is more knowledgeable about restaurants than anybody else I know. She’s the GRP food ninja.

And there you have it. UGC caters to the masses and isn’t “sliced” for us as individuals. I see a few companies at the edges building intelligence into their products including people like Lunch.com (share & discover opinions most relevant to you), Gogobot and Hunch (although I’ve never really asked Chris where he’s taking things longer term – maybe I should!).  I like his tagline a lot, “Hunch personalized the Internet.” That sounds right. If he can achieve that it will be big. Really big.

So either the future is a Lunch.com like sacking of websites like Trip Advisor (which given that he sold them his last travel venture for some serious bank sounds very plausible) or somebody developing technology to enable all of the existing products including Yelp, Zagat, RottenTomatoes – even Amazon.com – to be sliced better.

My Wish List:

I want at least three slices for any UGC sites.

1. Sliced by social graph – The obvious place to start is my social graph.  Sure, I’d like to know what my Facebook friends think about the wine, books, films, hotels, airlines, etc. that I’m thinking about using.  So I’d find this slice very handy. Necessary, but not sufficient.

A few years ago I planned an outing to Santa Barbara for my Dad’s 70th birthday. It was quite difficult to find the right place to stay. I needed somewhere nice enough that people wouldn’t complain about lumpy beds but not so nice that people would complain about the prices – it IS Santa Barbara after all. At the time I had young kids but my siblings didn’t. (now we’re in the reverse situation). So I needed a place baby friendly but not too family focused.

Ugh.

Our social networks are filled with our friends, sure. When we’re all 25, they’re all pretty much exactly like us.  The older you get, the more people you meet, the more diverse your friends become and even your lifelong friends diverge. My closest friend from childhood (and best man at my wedding) decided not to have kids. Believe me our hotel & travel preferences are different!

2. Sliced by influencer graph – The second slice I’d like to see is by the “influencer graph.” I’ve talked before about the power of Twitter for information discovery. If I want to know about any topic I can follow experts who like to Tweet & link to that topic and I expand my horizons. It’s why I like following people like Shervin Pishevar, for example. We’ve only met once but I love tracking what he’s into politically as well as his startup views. He’s one of those guys that everybody should follow. Beats only reading the NY Times everyday.

So what if I could know who were the “scouts” for all the hottest new restaurants in NY, SF or LA? It would be a sort of crowd-sourced Kelly Amoroso. What if I knew who wrote the best stuff for kids dining in cities? What if I knew who the biggest influencers were of independent cinema (for which I’m a voracious fan), the fashion guides for men’s clothing, the eco-travel wizards who span the globe with kids.  I’d love to know people to “follow” on websites to help expand my tastes the way that Twitter expands my news sources.

No, seriously. I’m stuck in a music rut. I still love the classic rock I grew up with, grunge from the 90’s and rap/hip hop from the 90’s and naughties (00’s). But I’m sooo out of date. I kept asking the guys who are like me but way more hip musically (like Ian Rogers) than I what to listen to. I need a better system of music discovery. But music IS probably the leading category at providing these types of services. We need them for every vertical.

3. Sliced by people like me – But mostly I want my recommendations sliced by random people I don’t know. That sounds counter-intuitive so let me explain. I want to know people who are just like me. They like similar food, they have similar politics, they’re in the same age & stage bracket, they have similar financial situation, watch the same movies, etc. In fact, this data set is probably totally different by vertical. I might love the same films as 50 people who for whatever reason don’t like the same hotels.

So my ask is topical slicing by people like me. Help me cut through the clutter. And I don’t want you all who are different from me to be bothered with my eclectic taste in films (I hate Hollywood blockbusters and I drag my wife to foreign films of shepherd herders in Iran or EVERY Mike Leigh film) or my poor taste in music.

I look forward to the death (or improvement) of Trip Advisor and all the other crappy UGC sites out there that have no intelligence or targeting whatsoever. I want Lunch or Hunch or whoever is going to solve this problem to hurry up and sack the incumbents and spill the blood of the outdated, crappy UGC Internet.

  • http://www.missi.com/ Peter Beddows

    So do I: Defines the challenge from practical experience and adds “flesh” to the issues outlined in Mark's post.

  • http://www.funnelscope.com/ kulsingh

    Phenomenal post, Mark. Thanks for the insights. As you suggest, data & analytics is key to making this happen. Over at Funnelscope we already have over 5 million data points for hotels that allows us to achieve these correlations. One slight contradiction I see in your post is you suggest that horizontal approaches could solve this yet there have only been successful vertical approaches (as illustrated with Netflix). And I don't believe my taste in movies correlates directly to hotels and books. But I agree opportunity is huge and game-changing. Love to discuss at some point when you are in NYC :)

  • http://www.missi.com/ Peter Beddows

    I agree with Mark Essel's response; love your answer, but I also see that you have added some real-life experience meat to Mark Suster's pragmatic definition here of the issues that we all are facing in our attempts at finding what “we want” when searching as distinct from finding what the crowdsourced “want” conclusion may be.

    Some time ago now, my wife and I needed to find a convenient motel close to Esher (suburb of London) in the UK: We searched and asked and and ended up at “The Monkey Puzzle”!!!

    Should have seen the name as a caution: Well named as a warning; never go there again but it came well recommended.

    We were woken several times the first night by false fire alarms and the next day went to breakfast in their dining room that was overwhelmed by the fowlest of odours of stale tobacco smoke and beer since the restaurant was also attached to a night-club on the ground floor. So we spent the first part of our second day driving around looking for an alternate.

  • http://www.missi.com/ Peter Beddows

    Mark Essel; I think your have defined the challenge underlying creating an effective, reliable, algorithm very well in your post about relevance. That is a great read in and of itself.

  • http://hdemott.wordpress.com Harry DeMott

    It's funny, I should never use Brits as an example of how I screen on Trip Advisor – but I have found it consistently the same.

    I feel I can take this liberty – as my wife and her family are all from Haslingden – which is a small town north of Manchester.

    It's funny – I stayed up north near this town in an inn that came well recommended by her relatives living near there – and had the same experience – awoken by fire alarms, placed smelled like the pub floor.

    Another trip we tried a different place and ended up actually having to break into the inn after hours. There was no key left where they said (they forgot).

    what “we want” changes all the time on our situation: our family situation, our financial situation, what we are traveling for etc…

    Same thing for restaurants – movies – etc…

  • http://www.funnelscope.com/ kulsingh

    Harry- great example! I don't think all these nuances can be solved programmatically and that is where the social element comes in- as you state: “Or is it a whole new system built to be social and malleable from the very start (I like this idea much better).”

  • http://www.missi.com/ Peter Beddows

    Harry, I can totally relate. We often use the reference to “The Monkey Puzzle” also as an example of how many Brit's appear to be unaware of how unappealing some of their “favorite quaint haunts” actually are to foreign visitors: This coming from – as I believe you know – me having been born/bred UK and very happy to now be a “yank” living with my born/bred American wife close to San Diego. What a blessing though I do miss frequently my old house and friends in Wales!

  • http://hdemott.wordpress.com Harry DeMott

    Funny. My in-laws live in Santa Cruz California (which is where my wife largely grew up) and while they love visiting home – you couldn't pry them out of California. No dark winters. No cold damp.

    Like I said in the original post, what always amazes me is the level of service demanded by UK travelers – particularly when they head to service light destinations such as the Caribbean. It's laid back – which is the point.

  • http://hdemott.wordpress.com Harry DeMott

    Interestingly, in my example – I do think that this could have been solved with a program. If you were able to screen on proximity to snorkeling, activities and restaurants – my hotel would have come up along with perhaps 3 or 4 more. Would have made it a lot easier.

  • http://www.missi.com/ Peter Beddows

    We used to live in the SF Bay area and know Santa Cruz well.

    Before coming to the 'States in '79, I was building a new business from Green-Field to full operation in Wales. We typically experienced 183 days and +18″ of rain, never mind snow, a year there. When I sent status reports back to the head office in Japan about weather delays, they got to the point of not believing it. Japanese chairman visited and his limo drove onto what looked like paved road but actually was remants of cold-dust mixed with mud. Car sank to axles; chairman had to roll up trousers to knees, and we had to fork-lift the car out. He said to me “Now I believe” lol

  • philsugar

    http://bit.ly/fzEItK

    It undisputedly had the most likes of any comment.

    And if you look I did keep my promise to blog.

  • bobfet1

    I think this is why Twitter's “interest graph” is one of the most under-utilized assets on the web right now. #1 on your list sounds like recommendations based on your follows on Twitter. #2 sounds like recommendations based on the Klout scores of people on Twitter (and possibly their relation to you). #3 sounds like recommendations based on the similarities between you and other Twitter followers (whether they are connected to you or not).

    In my opinion a lot of the problem with recommendation systems people are building right now is that they require too much work on the user's behalf – you have to rate a bunch of stuff to get recommendations. The good thing about Twitter is that you've been doing the work all along by accumulating follows all along – sites should just be able to plug into that and give good recommendations instantly.

  • http://www.BetterCEO.com John Seiffer

    How about doing this for politics? The problem with democracy is that all those people who don't think like me get to vote!!!

  • http://www.rentsavy.com Mike

    If Tripadvisor would publish the age of the person complaining (e.g. I'm in my 20's, what do I care if a 65 year old thinks my hotel in Cancun is too loud), my complaints would decrease by a factor of 5.

  • http://www.rentsavy.com Mike

    I love what Bizzy is trying to do. Sleek UI and the experience is really enjoyable as well.

  • http://www.rentsavy.com Mike

    I think that's how Groupon started (initially “The Point”) before realizing there's no money in it.

  • http://www.funnelscope.com/ kulsingh

    Yes I agree IF there was such a robust structured database around “proximity to snorkeling….” Our vision at Funnelscope is that this database can be developed over time but by first leveraging a social element. If you have input from folks / experts that can tell you “The following hotels are close to snorkeling” you would get your answer and then that data could be available to others in the future via programmatic means. I think this only scales with a process of humans first, machines second.

  • http://apsalar.com/ Ted Barbeau

    Woah. You just articulated what has been driving me bananas for the past couple years. Every time my lady-friend and I want to go out to dinner, we turn to Yelp. The problem with Yelp is that 80% of the restaurants seems to be between 3 and 4 stars. On an absolute level, there is probably a difference between the 3 star restaurant and the 4 star restaurant. But in reality, it seem like such a crapshoot to me and difficult to forecast. On the way to the restaurant I always find myself hedging like “Yelp says this is $$ and 3.75 stars…it should be a good meal…but who knows?”

    I'll take a friend's recommendation over Yelp/TripAdvisor/Netflix/ 100 times out of 100. But that is laborious and inefficient (and not always on demand).

    I only have a minute but I need to catch up on the comments and see what other people are saying later because I'm pretty passionate about this.

    Thanks for the post – I was turned on to you blog a while back* and have been hooked ever since.

    *I think by Google Reader's Recommended Items. Oh the irony!!

  • Andrew Perlmutter

    Given the selection bias in the types of people who actively post comments, I wonder whether the sort of granular slicing you want is currently viable. Are there really enough people “like you” or even “in your social graph” who regularly post comments across a variety of industry verticals?

    It seems that the success of the Netflix's recommendation engine is contingent on multiple like/dislike entries from millions and millions of users. But the primary reason why I rate movies on Netflix is because Netflix tells me that this activity will improve it's ability to recommend movies that I like. Unfortunately, the vast majority of sites where UGC's are relevant do not offer me anything in return. If I post a comment on TripAdvisor, it will not help me get a better hotel on my next vacation.

    So because the vast majority of internet users are not otherwise inclined to comment about the majority of their product experiences, we are left with a small circle of people making all comments. Doesn't this leave a very small probability that comment slicing will produce good results for you? Thoughts?

  • http://twitter.com/mgrosso Matt Grosso

    Mark, what might come closer to the service you want is item level collaborative filtering instead of user level collaborative filtering. Instead of finding items used by people like you it finds items used by people who also used an example item. When you are trying to branch out into something new like a new branch of music this method is easier.

    You may also want a popularity filter for when you are in the mood to explore something less well known.

    For example, on upto11.net you can submit or browse to an artist that you like; the algorithm there uses p2p sharing data to return 30 artists that were featured prominently in the p2p offerings of people that had the artist you started from. A slider lets you remove all artists above a certain level of popularity so you can see more obscure artists. (full disclosure: upto11.net is an experiment by myself and 2 partners in p2p thought crime from 2005.)

    For hotels, you would submit the name of a hotel similar to what you have in mind and an additional geographic restriction. This way you don't have to have the relevant users in your social network neighborhood or be anything like them to get the recommendation benefits.

    Item level filtering is a lot simpler computationally as well, since the prep time scales with the square of the item count rather than the square of the user count and the runtime look up is either constant time or the log2 of the number of items.

    The very real drawback is that user has to tell you what they are in the mood for, while user based suggestions can just look at your collection and suggest something new.

  • http://twitter.com/scrollinondubs Sean Tierney

    Mark, more accurately – you don't want a Pandora recommendation engine for unconstrained entities, you want a Last.fm-style engine. The difference being: Pandora = nature while Last.fm = nurture. Maybe this is semantics nitpicking but it sounds like you essentially want a way to do collaborative filtering unbounded by a particular vertical and have the ability to ad-hoc adjust the views of various results by tweaking the knobs on which relationships it uses to determine the suggestions. This is how Last.fm and other collab filtering-based recommendation engines work. Pandora uses qualities inherent to the media itself like rhythm, timbre, orchestration, etc (ie. the music genome) to suggest similar results. You definitely want a collab filter based approach.

    FB is in the best position of anyone to pull this off. They have the largest collection of vertical-unbounded like data combined w/ the ability to marry it w/ demographic and social graph info. It'd be a matter of them putting a sensible search interface together to allow people to slice on the dimensions you suggest. I'm sure they'll do it eventually.

  • http://www.skepticgeek.com Mahendra

    Mark,

    Nice post (as usual), and I like it more as Relevance is a key interest area for me.

    Back in July last year, I wrote about the following approaches to slicing for relevance:
    - Algorithmic Filtering
    - Filtering Based on Social Graph
    - Human Filtering
    - Crowdsourced Filtering
    - Shared Sources Filtering (Meta)
    - Influence Filtering
    - Social Search
    - Location Filtering

    Apart from UGC sites, I think relevance filtering is needed on many fronts today, including social streams on various networks. Also, I believe services that adopt flexibility in terms of which slicing approach you'd like are far more likely to succeed than those adopting limited approaches.

    If you're interested, see my post at http://www.skepticgeek.com/soc…/

  • Stijn vervaet

    Nice post again. I like what 22tracks.com does. An Amsterdam Start-up looking to scale the concept across Europe (and the world). 22 genres of music curated by people on the pulse. Brilliant way to keep up to date and discover new music.

  • Dave W Baldwin

    Great post! This hits on that e-mail sent showing a bigger diagram regarding what I'm after…mind you it was not spiffy, just showing the bigger picture:

    1) You have to have feed back regarding destinations/services and entertainment. The majority of us do NOT want to fill out a survey and that will not change. You have to gain opinion in a way most friendly to the user. If you were to ask someone what they thought of their trip, they probably would give a quick take on what happened, ranging from the motel to a restaurant to recreation and so on. It is a matter of understanding the random tidbits gained and sort them to their subject. This leads to a more real time source of data because you do have the info regarding the source plus opinion… it is a matter of sorting out if the person said the place sucked, too much this, that and the other thing.

    2) You have to sort the data quick enough to allow the intake (opinion) and disbursement (receiver) where the person planning what to watch, listen to, travel to knows what is happening now.

    3) If there is more information to gain, then it should be happening while the user conducts normal activities and can get an update regarding information.

    4) Places wanting your business need to get notice of interested customers and be able to send an invite IF their place fits the profile of the customer. At the same time, the customer's tool should be able to sort those places that truly fit and gain additional information.

    I could go on, but you get the picture. There is a way to do this that moves beyond travel sources that just flatter the advertiser and gets down to the meat and bones. Nothing against advertising at all, but you have to seperate that fluff from what you're really looking for.

  • Aleksander

    Very timely post. I'm actually working on a prototype of a larger concept where one of the main innovations would be a new approach to recommendations. I considered all of the things you mentioned in your wish list but after a chat with my AI professor decided to combine them with a method that as far as I know isn't yet used.

    Of course my current location isn't really helping, but it's only one more step I need to take to get my company started and on it's way. I find your blog quite inspirational and educational so thank you for that.

  • JoeYevoli

    I think social graph recommendation is something that is definitely under utilized. It's useful to see “the web's” ratings and recommendations, however, a single recommendation from a trusted friend trumps that entirely.

    If I see a restaurant that has a 3 star rating on seamless web, I'm probably going to ignore it. However, if I know that a trusted friend says “The menu's not that great, but the turkey club is the best in the city!” I'm absolutely going to get that turkey club.

    My social graph has been built, the next question is, “how can it help me?”

  • Marks

    discovery is over mark. i have nothing more to 'discover” in Boston.

    what i need now is contextual discovery.

    yelp and trip advisor jumped shark for me some time ago.

  • http://danreich.com danreich

    Love this post, very validating because these are some of the things we are currently working on at spinback.com.

  • http://www.Fashion-Ade.com Ella Dyer

    Hi Mark,

    Funny you mention a Pandora-like experience; we created just that for a woman's wardrobe. When consumers utilize http://www.fashion-ade.com/ they benefit from the recommendations made based on what they already own.

    Additionally, these consumers are declaring what they own, use and share with their social networks.

  • http://twitter.com/skaragiannis Sotiris Karagiannis

    So, what you need is to bring Ai into the game. Yes, AI. Because the matching process is going to be extremely intensive for traditional computation and databases. AI on the other hand, has been evolved the last 20 years (since I was in Edinburgh university studying Expert systems and Neural nets) and can now safely be deployed in such projects. You have taxonomies and the challenge is how to analyze such big data sets and deduce personalized attitudes from every single small and otherwise irrelevant piece of data. Of course, the percentage of hits will not be high in the beginning, but AI techniques allow easier correlation and self learning processes as well better behavioural predictions. Now, what platforms are you going to use and what paths are you gooing to follow ? In AI, each problem has each own way of solving it – in terms of AI algorithms or trends. It is mostly a R&D area, but if you find some University papers that are applying Ai into social media data, that might be the way to go. the real magic word here is ‘Social’, because social by its etymology, brings the concept of people making groups based on moreorless thei common criteria of what is cool, fun, good or bad. You do it in your life, hang out with guys or girls that you share common perception of things with them than with people you do not..

    I would say : Go ahead and researh this point, There has to be a solution, not obvious but there surely is. After that, everything will be easier.

    Good luck.

    Sotiris

  • http://wikidi.com/ Michal Illich

    Well, I think number 3 can be handled only by Facebook. And it's relativelly simple for them
    - they have a huge graph
    - the algorithms for discovering entities with the same properties (i.e. people like you) are known (only they are very computationally hungry)
    - with their recent take on Foursquare+co they'll have a database of places (restaurants, hotels)
    - database of music is publicly available with open source licence
    The way is known – the question is only if and when they will do it.

  • paige craig

    Totally agree with you Mark – with millions of people joining the online world every day this is a serious problem that gets worse every day. From inbound emails, customer service requests, ads, search, purchasing and even dating we are all struggling to find the signal in the noise. And yet I think Klout is poised to solve this problem (NOTE: I am a biased investor). Personally I use a pretty strong friend filter in everything I do to cut through the noise – Klout takes that concept of the friend filter and has extended it to work for everyone. Klout can solve this problem by providing everyone (call centers, email hosts, ad agencies, UGC sites) with a consistent and standardized influence & relevance score (i.e. your Klout). I'm waiting for a world where my hotel decisions, inbound emails, targeted ads and even my dates are filtered and aggregated by the Klout wizards

  • Carlos

    check out http://www.kikin.com … we are getting there …

  • http://www.facebook.com/profile.php?id=1302811786 Joris Van den Broeck

    ” To make the recommendation engine work well you need to register and build your profile so that the service provider has a chance of knowing what you like and can provide you with good recommendations. ”

    I don't think that's true, these days you can get tons of information for free; you can analyze the users
    * twitter feed
    * facebook stats
    * previously booked trips/hotels
    —— if there're tons of comments of “bad service” on the hotels you books from other people, the system can learn that you don't care about “bad service”

    Ofcourse, it won't be easy, combining all the information, knowing what's relevant and irrelevant, but I'm shure it can be done. I'm going to think more about this idea, I'm working a lot with AI these days and think it's an interesting challenge:)

  • http://twitter.com/skaragiannis Sotiris Karagiannis

    Actually what msuster needs is globalhabit :) But anyway, in the correct path.

  • Sol Eun

    I totally agree with you that we need to be able to find reviews those are more relavant to us.
    I have been working on this problem for awhile and I believe I have a solution.
    Reputation graph from your peers and people with similar background would be able to suggest more precise reviews/recommendations.

  • http://twitter.com/quotify Ben Ross

    Girish’s comment is on point. Word of mouth is very powerful, but we believe that the answer to this problem is to use the power of data and an algorithm to match me to the right service provider.

    We have made a business of matching consumers with service providers by developing a consumer matching engine which powers a lead generation business in Australia called quotify.com.au (like ServiceMagic – but in Australia).

    By mining the consumer feedback data we gained from matching 100,000’s of consumers with small businesses in Australia we developed a taxonomy of the key attributes which matter to consumers for each type of businesses (eg. timeliness and price for plumbers vs courteousness and quality of service for funeral directors).

    We then understand as much about consumer preferences as possible by leveraging cookie data and also by asking consumers some questions in a similar way to hunch. For example, we can potentially deduce a lot about your likely preference for a plumber by knowing your favorite movie, therefore in our forms we can ask questions that may seem oblique (or fun) to the consumer but actually provide us with significant insight into consumer preferences.

    With this data, we segment consumers into certain categories (based on their preferences) and have found the patterns for when the certain types of consumers are most satisfied when matched to certain types of businesses

    This high quality matching has increased the value that we can sell each lead for to the right service provider.

    The platform is called http://www.quotify.com. Thanks for the great post!

  • J Dunitz

    There is an active and growing early-stage site (approx 1 million uniques per month) called RANKER.COM that provides an excellent antidote to the useless numerical rating and recommendation sites. The RANKER site is a repository of user-generated LISTS on all topics.

    Ranked listings of objects (places, activities) offer an efficient snapshot comparison of multiple options within a category — which is significantly more evaluative than a review of a single object (place, activity) in isolation. In addition, because a list includes multiple entries, the list-maker is revealing clues about his own milieu, thus providing a context for the reader to assess the “validity” or relevance of these suggestions to the reader's own proclivities. In this way, lists are enormously more helpful than one-off commentaries.

    RANKER also posts CrowdRanked lists which are master lists of any particular topic derived from aggregating all lists on a particular topic based on software that assigns weighted values to discrete items within all such lists. This is RANKER's answer to 5-star rating systems that just tally votes. The master lists offer much more nuanced preferences of the crowd.

  • http://www.johnexleyonline.com JohnExley

    Mark, this is exactly exactly exactly one of my biggest visions for what I hope the future of the web will look like. Love this post even more than usual.

    I recently wrote on my blog about this sort of problem with the 'social travel space'…I wish there was a company out there that somehow married Hunch's personalized recommendation technology with say, Dopplr‘s extensive knowledge of cool things to do in cities around the world.

    Don't want to spam your comments, but here's the post I wrote a couple wks ago if you're interested:
    “The Problem With Planning A Trip To NYC…Or Any City For That Matter” http://johnexleyonline.com/201…/

    Thanks!

    - John X

  • joeagliozzo

    Seems like Netflix/Amazon style recommendation engine would work – “people who bought/rented ___ also bought/rented ____”. Once your purchases start to match the pattern of someone similar you get recommendations on new items. Question is how to automate collection of transaction or buying information since you don't control it through the site – maybe “Mint” style through credit cards? Could you get customers to buy through your new “expedia” style site with the promise that as they book more hotels they get better and better recommendations? but how to get them some immediate gratification to encourage uptake? Hunch style questionnaire? Demographic questionnaire?

  • http://www.facebook.com/bdclimber14 Sean Coleman

    I'd like to build a generalized collaborative filtering service that cross-recommends the rich, action-based profiles of verticals like Pandora and Netflix. Imagine recommendations from not just other individual movies based on other movies, but on entire abstracted profiles.

    For example, if your and my Netflix “activity profiles” were a match, then your music selection may be of interest to me. This could be particularly useful among verticals with thin data. Generally, people watch movies and listen to music more often than you book hotels. If there was a crowd of individuals all with similar music-listening, movie-watching and restaurant-dining tastes as you, then their hotel interests could be a very strong signal, more-so than the general crowd, or sub-culture of Yelp.

    I think leveraging cultural verticals like music and movies could be the mechanism to achieve 3.

  • http://www.elieseidman.com Elie Seidman

    Chiming in here since the problem you face is the problem we've been working on. I don't think we fully agree on the tactics of how to solve the problem. We strongly believe in the role of the trained expert who knows how to evaluate something based on the needs YOU, based on your need group, not where you grew up etc, have.

    If you have not already done so, use Oyster to actually plan a real hotel stay and trip. I think you'll find – as our customers have – that it's the best thing out there by a lot. Faster, simpler, richer, more expert, etc. Our current limitation is that we have the 14 largest markets in the US and Caribbean but are missing the 10s – if not 100s – of other markets our customers want to go to. We'll be adding another 80+ destinations in the months and quarters to come so that problem will largely be solved.

  • http://www.elieseidman.com Elie Seidman

    Oh, and if you want advice on Orange County hotels, I've recently been to the Montage, St Regis (Monarch Beach) and Ritz (Laguna Niguel). They'll be on Oyster soon. Let me know what you're looking for and I'll give you some advice plus send you photos.

  • petegrif

    Absolutely right. I have a similar idea which I call 'microworlds.' People say things like 'How can you live in LA, it's so big?' But they miss the point that none of us really live in 'LA' – we live in a microworld of our own construction. A person's microworld is a set of preferences – friends, restaurants, schools, movie theaters, stores… What matters is not how big a place is, or how small it is, but whether can support the kind of microworld I need. What you are looking for with your 'recommendation slicing' is the ability to quickly create and/or evolve such microworlds. And you are 100% right that most existing tools don't help.

    btw Mark – happened to meet 2 ex colleagues of yours at Founders Showcase in San Fran tuesday. Gave you outstanding recommendations for your work at salesforce.com. Good to know.

  • http://twitter.com/patmccarthy Pat McCarthy

    Wrote a blog post covering this since it's such an important topic, and there's some great comments here about how it's a difficult challenge to achieve all three slices.

    I think it's wise to take the approach of nailing recommendations in one slice at a time in building out a service for this, and the social graph is the right place to start for a few reasons:

    - Requires less work from the user
    - We may not share tastes perfectly with friends, but we know how much we can trust each one of them and filter their recommendation through that lens.
    - Friend recommendations have serendipity, they create conversation, and they are realtime (not stale old review data).
    - Allows us to leverage the Kelly Amorosos of the world.

    There is a ton of value in the other slices as well because the social graph does break down at certain points. None of the three slices is perfect, so the right service will need to figure out how to blend all of them well to create an experience that works and doesn't require too much effort.

  • http://yourplace.com Brian Wilson

    I am with Paige here. I think Klout is positioned to do something very powerful. The concept of “page rank for people” is brilliant. However, they need to provide Klout scores based on subjects / verticals for it to be useful.

    A company who can create the “credibility” standard by vertical can build an API that all community sites will want to use in order to intelligently promote the contributions by the subject-matter experts. This was the initial glow of Quora right? You can ask a question about a company and actually see an answer from the CEO. Awesome.