• http://www.everfluxx.com/ Everfluxx

    Excuse me Danny, but where in his post did Amit Singhal say that Google has started using merchant reviews as a ranking signal?

  • http://daggle.com/ Danny Sullivan

    Good point — I’ll get this clarified. I’ve made that assumption because it talks about not using reviews for display purposes, which I read to imply they were used for ranking.

  • http://www.webtraphic.com/ Brian Cox

    I still see ‘said merchant’ for some pretty large (money) terms like ‘Discount Designer Sunglasses’. I’m pretty sure he’s down 1 spot from where he was yesterday though.

  • http://www.webtraphic.com/ Brian Cox

    correction, that was with personalization.. hes gone..

  • http://www.everfluxx.com/ Everfluxx

    Thank you very much for your clarification, Danny. Obviously Google will never disclose exactly how they construct their “opinion” of the users’ opinion, so I agree with you that it is perfectly legit and reasonable to assume that merchant reviews are likely part of the new algo (since they did not explicitly deny it, either).

    Speaking of signals, as we know ratings are much easier to treat algorithmically than reviews: no sentiment analysis required there; it is just a matter of counting stars or weighing points on a Likert scale. However, user ratings are also trivial to spam, unless one has to actually complete a transaction in order to be able to rate a merchant. So, I believe that ratings collected through Google Checkout’s seller-rating system might be a cleaner, more effective signal source.

    My hunch is that Google might also have thrown an entirely new source in the mix, one that -AFAIK- they admittedly have not used for rankings until now: i.e., conversion data gathered by Google Checkout and/or (maybe) Google Analytics. That would be an excellent source of implicit customer satisfaction signals, quantitatively and qualitatively speaking, because what Google improved, as Singhal said, is their ability to demote merchants that offer a very bad overall user experience. Think about it: a disgruntled customer might or might not leave negative feedback for other customers’ sake; but most likely, they will never purchase anything from the same website again. What do you think?

  • http://daggle.com/ Danny Sullivan

    I thought about them using Google Checkout or Google Analytics data, too. The problem is that not all merchants use this, so it’s not a complete solution. They should use toolbar data and things like bounce rate to detect dissatisfaction. But that wouldn’t work well, either — after all, Decor Your Eyes had people who were converting. Sure, maybe they don’t go back — but there are plenty of merchants I don’t go back to. I bought my new eyeglasses frames about a year ago. They’re still fine, so I have no reason to go back to the legit merchant I used back then any time soon.

  • http://bandago.com Sharky Laguana

    So what are they going to do to prevent competitors from posting malicious false reviews and negatively affecting page ranking?

  • ippy

    Google says that links in NYT contributed to the site’s ranking. I didn’t see any links at all in their article and I actually looked for them. Not sure about Bloomberg..

  • http://www.Weberest.com weberest

    In my opinion if Google starts using reviews as a ranking factor will be wrong. As a company grows bigger it will have more and more customers. The probability of having dissatisfied customers gets bigger and we all know that people write reviews when they are dissatisfied and a lot less when they are satisfied.
    My point is if Google starts counting bad reviews, they should take the number of customers into account.

  • http://www.everfluxx.com/ Everfluxx

    Danny, on second thought I think I agree with you: even zero returning customers over a long period of time might not necessarily imply customer dissatisfaction for merchants selling durable goods, so if conversion data are being taken into account, the resulting signal is probably not very strong. Post-purchase seller ratings gathered by Google products and third-party services such as Kelkoo are probably the best signal source.

  • http://www.stuartbell.co/ Stuart Bell

    There is a big risk of malicious negative feedback reducing ranking, especially in the local market where numbers are smaller. The yelp model may be an indication of approach, where one time a ‘yelper’ is discounted over a regular reviewer. Again, not fool proof though.

  • http://www.michael-martinez.com/ Michael Martinez

    You’re doing a great job of heading off a new SEO myth at the pass, Danny, but it seems that people are already sharing the New York Times article in forums (other than Sphinn) and not checking the links to see if they’re nofollowed.

    This issue could inflate into a 2-year-long fight against ignorance, as has happened in the past. It’s a shame the article didn’t mention that the links were nofollowed to begin with. Maybe that would have stopped people from drawing the wrong conclusion.

  • silence

    Danny, in your screenshot to “chanel 5117 sunglasses” search results, the second website eyeweartown.com belongs to Vitaly Borker, as well :)

  • http://SEOinKorea.com Korean SEO Guy

    Thank you for your good description.
    RE Google algorithm, firstly I thought PR algo. tuning. There will be a little possibility.
    But, this method have problems, too.
    Finally, the possibility is in their inner services such as analytics, webmaster tool,
    Checkout, etc.
    At there, if they apply “critical data” concerning sentiment analysis and integrate with
    ranking algorithm, G will be control SERP.
    Already most of sized US e-commerce sites use these services and the sites gaming at SERP 1-10 pages are these sites.
    In the end, this tuning will not bring out distorted SERP.