Turing YP Sites into Recommendations Engines

I wrote very briefly yesterday that CityVoter’s redesign and new direction was ahead of an emerging trend toward real time recommendations for local. Previously, I said the following about Twitter and Google:

Online I can go to Yelp or Citysearch and look at consensus views and ratings. But in my vision of Twitter’s future I simply query my Twitter network and I get a bunch of responses to the lunch recommendation question. And I get them more or less instantaneously — or in “real time” if you prefer.

There’s also this earlier post, Twitter, Vark & ‘Real-Time’ Local Search:

Not all these services are the same, but conceptually the idea is to leverage human beings to respond to specific questions or queries either in real time or in a near real-time way.  Social search, review sites (e.g., Yelp) and Q&A (e.g., Yahoo Answers, Askville, LinkedIn Q&A) are all versions of offline word-of-mouth recommendations.

The promise of “human-powered search” has been around for several years. However none of the sites promoting that concept have really been successful. We’re just starting to see something more viable crystalize and emerge, in all these sites, which may well represent a successor to traditional search — or perhaps a companion to it.

And in October of 2007 over at Local Mobile Search, I wrote about a service called Mosio (as part of a category we were calling “social directory assistance”):

It can be viewed more broadly as the prototype for the sort of social/mobile/search service destined to be the foundation of a range of local search-based businesses.

Sebastien Provencher yesterday wrote a post introducing Praized’s “newsfeed, real-time search and conversation platform.” It appears to bring together a range of social media tools to YP companies, resulting in a buzz/news feed and Q&A functionality. Yellow Pages Group has apparently implemented part of this in the form of “Yellow Pages Answers“:

Picture 4

Here’s how Sebastien describes how the service works:

Answers (a “local” Question & Answer service, including a social network broadcast mechanism). Consumers can ask questions to the community and to their Facebook/Twitter friends and all answers come back to a unique page. Merchants can even join the conversation!

Inspired by Facebook and Twitter, it appears that Praized has found a very strong new direction for itself. This is quite powerful, but let’s step back for a moment and look at this functionality in a broader context.

The ability to ask questions of a group on local sites is nothing new. Both InsiderPages and Judy’s Book offered Q&A functionality from near inception (in 2007). It’s no longer on InsiderPages, now owned by IAC. And Judy’s Book has been sold and relaunched. Yelp has had questions and answers for several years as well (though not well integrated into the experience). Trulia and Zillow have very robust communities and Q&A tools in which local real estate agents can participate in conversations or provide answers to consumer questions. Beyond these there are a wide range of pre-existing services, Yahoo! Answers, Amazon’s AskVille, LinkedIn and several others, that also offer Q&A — though not in “real time.”

Picture 5

So what’s new:

  • The “real-time” (near real time) dimension of Twitter and now Facebook Status updates. (BTW: IM could have done something like this years ago but didn’t)
  • The ability to tap into multiple networks simultaneously because of APIs
  • Perhaps most importantly, Twitter has captured people’s imaginations and helped put a name/label on this phenomenon (“real time” search)

The underlying consumer behavior is simply asking for word of mouth recommendations and is as “old as the hills.” But the ability to efficiently ask many people for advice or a local business referral at once online is new. Reviews were step one; the combination of quasi-real time answers and social networks is an evolution of that phenomenon.

The injection of merchants into the conversation is useful and potentially powerful but something that Trulia, Zillow and Yelp were doing already and which Yelp is now doing more directly. There are other examples out there as well.

Implicit in Sebastien’s post are some very interesting suggestions for the future of yellow pages, which is part of a much longer discussion than I want to get into now. Make no mistake, the integration of Q&A into yellow pages and the ability to tap into other networks are great product enhancements. Conceptually, however, none of these things are truly new, just the packaging and presentation.

21 Responses to “Turing YP Sites into Recommendations Engines”

  1. predictabuy Says:

    Getting the packaging and presentation right is going to matter. Arguably, that’s all Apple has done with the iPhone and the iPod before it. When someone finally nails the right combination of ingredients it will probably seem obvious after the fact.

  2. Greg Sterling Says:

    Agree. Not trying to dismiss that. Your Apple analogy is exactly right — it’s the “user experience.”

  3. Thinking holistically about local search « Predicting What Consumers Want to Buy Says:

    […] of several new products promoting this shift – which Greg Sterling reflects on in this post.   And I agree with Greg that in some ways we have almost come full circle: The underlying […]

  4. Turning YP Sites into Recommendations Engines « LocalLab : Foire aux Infos Says:

    […] Small Business, User-generated content. You can follow any responses to this entry through the RSS 2.0 feed. You can leave a response, or trackback from your own site. Commentaires […]

  5. Frank Says:

    Interesting, when I read the title of this article I was expecting something very different. To me a recommendation engine is a system that uses analytics to actual predict how well you will like something (e.g. Amazon or Netflix). User-to-User recommendations have their place for sure but I believe that local search sites, especially those with the millions of ratings and reviews can do a much better job at providing true prediction and recommendation. To me the problem with asking the “web” for advice is you’ll always get back contradicting answers and will often have little context to help distinguish which advice is most suitable for you.

  6. Buzfactor Says:

    The nexus of real-time search and social networks…

    While the search interface today is simple, powerful and effective, the “one size fits all” approach is beginning to show its limitations. Moving forward, I believe our personal network of relationships will allow us to organize and filter …

  7. Greg Sterling Says:

    Frank:

    Sorry to confuse. Amazon style recommendations would work for events but not for service businesses.

  8. Frank Says:

    Greg,
    Why do you say it won’t work for the service business? If you take restaurants, I think the “amazon” style recommendation is very valid. People’s tastes in restaurants is something you can model relatively easily with good data. There are definitely patterns where people who like restaurant A tend to like B, so why not recommend B to those who rated A but haven’t rate B? That’s just a very simple example but I think it gets the point across. The algorithm we use has shown to have better accuracy in predicting peoples ratings of restaurants than Netflix does in predicting movie ratings and we have FAR fewer ratings (data points).

  9. Greg Sterling Says:

    Yes, works for restaurants. Not for plumbers, roofers, etc. That’s what I was really referring to. In the airport so doing quickly

  10. Perry Says:

    The other dimension of “conversational techniques” is that there is a chronic lack of online content addressing the myriad of “other considerations” which impact consumer purchases. Think of availability/inventory questions, ambiance, style, price policies and specials, insured/bonded features… All these lead the innovation towards conversational tools where leads can be qualified/directed in ways that reco engines can’t practically contemplate. Comparison and qualification will be conversational. The content needed to fulfill this is far beyond what is practical for us to expect to be available on the web for “engines” to process.

  11. Greg Sterling Says:

    Perry:

    Agree, recommendations engines/Q&A can be myopic or limited. But there is some new thing on the horizon that takes “conversation” between people about products/services/experiences and combines that with underlying content or data not generated in that moment.

  12. predictabuy Says:

    Greg and Perry – I think you are underestimating the potential for utilizing collaborative filtering techniques in local. I’m not suggesting its an answer in and of itself, but I’m astonished that it hasn’t been used at all (and aim to do something about that ;-)).

    One example. It’s often difficult to characterize some of the intangibles Perry identifies (things like ambiance). Sometimes its easier to infer this information from people’s actions — or what they actually say and do.

    There’s some interesting insights in to this coming out of research done for the Netflix Prize. Using ‘facts’ about the movies does absolutely nothing to improve ones ability to predict people’s ratings of a movie. If you’re interested, you can read more here: Netflix Prize has lessons for local search http://bit.ly/4fD7s

  13. Frank Says:

    Predictabuy – I wouldn’t say no one is using collaborative filtering, my site, theSUGGESTR.com, uses CF and a few other techniques to do “blind” prediction based on user similarities / item similarities and a few other special ingredients 🙂

    That said – I’m not aware of anyone having major success with it – we’re an unfunded site that was built in spare hours here and there – while the engine part is very accurate we haven’t had time to work on the messaging and presentation enough as we should – and we haven’t invested in marketing (yet).

    Also in our research we’ve found a few other sites that do stuff similar using CF – none of which have really stuck out as being a great solution.

  14. Greg Sterling Says:

    Not sure re the technology (whether it was CF) but BooRah was doing something with recommendations and previously Zync was also going to be a recommendations engine (bought by uLocate).

    TrustedPlaces in the UK has got a recommendations algorithm going (and there may be an element of CF there too).

    I think it’s a useful tool in some contexts, but I would say it won’t work for local across categories. Also, because of my diverse interests Amazon’s recommendations for me are often confused and off base — sometimes they work well.

  15. Frank Says:

    Yep, a few sites do it, all to varying degrees. My biggest issues with the sites I’ve found is that the algorithm is TOTALLY hidden. So you’re never sure when its being applies or when the recommendation is based on something else (location, sponsored link, date relevance, etc). One of the things we’ve tried to do is provide some context to the question of “WHY” a recommendation was made so that a user can determine its validity.

    Greg, I agree that this type of recommendation system is more useful for places like bars, pubs, nightclubs and restaurants than local search in general.

    Thanks for the great discussion on this topic – it’s nice to see it get some “airtime” 🙂

  16. Greg Sterling Says:

    I think that recommendations or some form of “push” or “discovery” element is valuable in LS. And certainly it applies to travel as well. It complements conventional search.

    There might even be a greater argument for this in a mobile context given the challenges and limitations of search on the small screen.

  17. SearchCap: The Day In Search, May 11, 2009 « Seo Costa Rica Says:

    […] Turing YP Sites into Recommendations Engines, Screenwerk […]

  18. YPG Promotes Answers « Screenwerk Says:

    […] Promotes Answers By Greg Sterling I previously posted on the new Answers service from Yellow Pages Group, but a formal press release went out yesterday: […]

  19. Turing YP Sites « LocalLab : Foire aux Infos Says:

    […] know this was a typo, Turing YP Sites into Recommendations Engines at Screenwerk but I wouldn’t hesitate to say Greg is not only purposefully but accidentally […]

  20. australian iphone apps Says:

    Thanks for the nice post! I think that recommendations or some form of “push” or “discovery” element is valuable in LS. And certainly it applies to travel as well. It complements conventional search.

  21. Brentwood Roofing Contractor Says:

    Thanks for the article Greg. This would be smart on the part of YP’s, as they missed the whole internet boat compared to the search engines. That could be a big change, but the problem would be the visibility to the actual consumer without a popular platform to get in front of the people. And with the SE’s killing them visibility wise, that the trick. But smart nonetheless.

Comments are closed.


%d bloggers like this: