• http://DirectMatchMedia.com Ben Cook

    Danny, while Rand never said Shares would improve rankings, they have heavily promoted their correlation data which as you mentioned, is often interpreted as cause.

    Personally I’d much rather have SEOmoz dedicate their time and effort to tests that show results rather than simply illustrating a fairly low correlation (the highest factors are in the 0.2-0.3 range).

    That being said, Rand does seem to be contradicting something Matt said.

    In this tweet (https://twitter.com/#!/randfish/status/78265937895239680) Rand says Google is clearly consuming the Opengraph+share data while Matt says Google doesn’t see the share data at all (according to this post).

    Unless I’m missing something only one of those two statements can be accurate.

  • RustyS

    A causation link may perhaps still explain the correlation, but more likely in the other direction. Higher ranking content will be seen more and likely will be more often shared. So rank might be influencing shares.

  • http://www.seomoz.org Rand Fishkin

    Hi Danny – thanks for the article and clarification. Matt’s statement was curious in light of this – http://min.us/mvjgaxK – anyone who’s logged in to Google can see sharing annotation from Facebook (you don’t even need to connect your Facebook account).

    In addition, I think Matt did not specifically say they don’t use OpenGraph data, which is where the Moz correlation stats come from. I’d agree it’s possible that Google doesn’t use Facebook data via the OpenGraph, but… well, it seems unlikely. I think a lot of SEOs are going to be running tests of URLs they share on Facebook, and I suspect we’ll see something.

    All that said, it’s possible Facebook shares simply have a very high correlation and no causal impact, but at such high numbers (and even controlling for links), it’s weird. I think it begs the question of why.

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

    These “correlation studies” don’t show that any scientifically valid correlations exist between anything.

    That said, two correlated events can share the same cause without either being the cause of the other.

    And all THAT said, these “correlation studies” are not looking at the factors that search engines use to determine rankings — they are just compiling statistics about attributes of Websites that happen to appear in the search results.

    To get to the ranking factors you have to be able to corroborate the “factors” you test for — and these tests are not designed to do that.

  • Judah Macabee

    On Twitter @randfishkin is posting screen shots of search results to prove his point, but he’s not proving any point at all. Reason he’s not is because his screen shots are of search results from real time search. I can find no evidence that Facebook shares are showing up in the traditional Google search.

    The real time search is not the traditional Google search so for Rand to say that “there’s a high correlation between Facebook Shares and ranking well on search engines” does not seem accurate.

    Rememeber that he runs a company selling people like us tools to rank better. If Rand’s assertions are true or if he sells them as true than he sells more products.

  • http://www.sunny-dessous.de/Impressum.html Sunny -Dessous.de

    As far as I remember from my math lessons a direct correlation has a factor 1. We here see a factor of 0.25 . That’s not a very strong correlation. So I support, what RustyS says. High ranking causes more views, getting shared more often.

    On the other hand I can imagine, that shares increase traffic, if not links as well, a metric that could influence Google rankings…

  • Julian Grainger

    Google has confirmed in the past they look at the influence of a profile within a social network. They have stated “We treat links shared on Facebook fan pages the same as we treat tweeted links. We have no personal wall data from Facebook”.
    http://searchengineland.com/what-social-signals-do-google-bing-really-count-55389

    They have also stated that if Facebook doesn’t give them the data they can get it from other people. This can only mean they are accessing social graph data from internet providers in the same way as Media6Degrees, or in similar vain, Hitwise does for traffic data.

    The correlation from SeoMoz will be indicative of the probability of the content being shared by influencers. In other words, the shares are not the cause but a good measure of the strength of the vector to influencers. And in that it is useful data.

  • http://www.mindshareworld.com Ciarán Norris

    There’s also the fact that Bing absolutely is using this data. It doesn’t mean much in Europe where they have no share, but surely this is worth considering in the US?

  • http://www.nuttakorn.net nuttakorn

    I think Google could index partial of Facebook Profile Share, you can try by using this query :
    site:http://www.facebook.com/permalink.php? , it returns you 139,000,000 results, i’ve checked many of pages are share from Facebook Profile not Facebook Fan Pages.

  • http://ilian.i-n-i.org Ilian Iliev

    Awesome post and explanation of the problem. Do you have stats about how much Google Realtime Search is used compared with Google Search.

  • http://www.imagefreedom.com/ Matthew C. Egan

    I think ultimately Julian summed it up perfectly, “The correlation from SeoMoz will be indicative of the probability of the content being shared by influencers. In other words, the shares are not the cause but a good measure of the strength of the vector to influencers. And in that it is useful data.”

    Shares are a great indicator that the content is worth talking about, and if it’s worth talking about, it’s worth linking to. Likes on the other hand, or links posted on Fan Pages, or whatever, is still up for some debate because we know Google scrapes SOMETHING from Facebook, but according to what Rand has tested thus far he’s not able to support that a Facebook share is as valuable as a Tweet has proven to be.

    You tweet something, it’s up there in a matter of hours, but his Public Fan Page Share test has thus far proven ineffective. I only find his Facebook post on Google, not the link he posted.

  • http://davelawlor.com Dave_Lawlor

    I can tell you that from watching the twitter stream coming out of #SMX that an overwhelming amount of people tweeting did take it for a fact that correlation is causation, it was by far the most prevalent tweet I saw from the stream for a good amount of time that Facebook shares promote ranking. I cant imagine that so many people would take it wrong if it wasn’t presented that way.

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

    Dave_Lawlor: “…an overwhelming amount of people tweeting did take it for a fact that correlation is causation.”

    Captain Barbossa: It would strain credulity at that.

  • http://davelawlor.com Dave_Lawlor

    @Michael Martinez I probably worded that wrong myself, meaning that most of the people tweeting from that session really weren’t looking at the root data they just took a “take away” from Rands presentation

  • http://www.TradeDoubler.com Nate

    Given the confusion amongst the general audience, I’m questioning the wisdom in releasing causation data without an expert causation overlay. Interesting how the SEOmoz authority has the second highest correlation. Funny that.

    I think the study is good, and the SEOmoz team are producing some great stuff, but it might need a little filtering. Still, a little confusion does generate a good amount of buzz and backlinks, no?

    Agree with Sunny -Dessous: the Facebook shares could be triggering site visits and content usage metrics that will never show up in a Google SERP.

    Nice coverage of the discussion Danny, really enjoyed it.

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

    Dave wrote: \…meaning that most of the people tweeting from that session really weren’t looking at the root data.\

    I understood what you meant. However, most people in the SEO industry don’t habitually question everything they hear from SEO conference panelists. The majority of us probably wait for some sort of corroboration. But it usually takes some Captain Kirk in the back of the room to ask, \Why does God need a starship?\ before people wake up and start to say, \Huh?\

  • http://www.VerticalMeasures.com Arnie K

    Just want to chime on two points.
    1) Danny amazing piece of work considering you have a conference to run and you did this so late at night.
    2) Ben – “a fairly low correlation (the highest factors are in the 0.2-0.3 range).” Sorry but using Spearman, a 0.3 correlation is actually quite strong.

    Arnie

  • http://www.fangdigital.com Jeff Ferguson

    Danny, I don’t think the point of the backlash was that one side was right and the other was wrong, but instead that the correlation studies are done at all.

    Per my blog post yesterday (http://fangdigital.com/2011/06/07/i-dont-care-what-you-think-about-seo-ranking-factors/):

    “SEOMoz* likes to beat the drum of “correlation does not equal causation” as a way to keep people from overreacting to these results, but that’s just not the way humans work. Our attention spans last about as long as it takes to get the disturbing data and drops off long before the warning not to panic.”

    These studies run rampant SEOs use them in their optimization when they shouldn’t and they end up doing real damage to their sites or waste time on items that don’t make a difference (in general terms, in this case, getting Facebook likes and shares isn’t a bad thing in the name of social media marketing, but it shouldn’t be done under the guise of helping with SEO).

    As the title of my post says, I don’t care what others think affects Google’s rankings… I care what Google knows.

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

    Annie, any value between 0 and .5 is considered a “weak correlation” for Spearman. 0 to .25 indicates no significance. 0.25 to .5 indicates *some* significance.

    The SEOmoz data doesn’t prove much in the way of correlations and certainly isn’t very relevant to RANKING FACTORS.

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

    Sorry, Arnie. I misread your name (I blame the font, not my eyes).

  • http://www.VerticalMeasures.com Arnie K

    Hey Michael, no problem on the “Annie”. I respectfully disagree on the correlation values, but totally agree that this data does not PROVE anything. I think the only thing the Facebook shares proves is that good content gets shared. And good content gets links.

  • http://DirectMatchMedia.com Ben Cook

    Danny, thanks for the postscript on this post.

    I think it illustrates at least two important points.

    If the blog coverage for Bruce Clay took Rand’s presentation as saying that Facebook Shares help rankings, it seems reasonable that a large majority of the people listening took it that way as well.

    At the very least, I think SEOmoz needs to be MUCH more responsible in the way they present their correlation data. Saying \Correlation is not causation\ but then speaking about the data as if it were, doesn’t cut it.

    Also, as I mentioned in my first comment and as discussed in the post script, it appears there needs to be some further clarification of this issue because some of what Rand is saying most definitely contradicts what I think we all understood Matt to be saying. Given that, it seems like it actually is a case of \he said, he said\ at least in terms of what Facebook Share data Google has.

  • http://www.tinyurl.com/ivan-temelkov Ivan Temelkov

    Some great controversy here indeed. I know that Google isn’t fond of Facebook or its practices although apparently Facebook Shares have correlation and reflect on SEO visibility. Thanks for sharing the news from SMX guys.

  • http://www.hortonwebdesign.com Chris Horton

    Wow. Yeah this is all confusing and really doesn’t seem to prove much at all other than Google does see activity on Facebook. One should also consider that if you search Google for a term like “bussangle nefaribook”, is there something in the algorithm that says 1) is this a misspelling, a) if yes, provide alternative, b) if no, is this something we don’t know about? And if it’s that “b” option, does it then go into Social Media overdrive attempting to locate other people that are using that term in Facebook, Twitter, Digg, etc.

    All Google cares about at that point is to find some type of relevant content to display no matter where it finds it. It probably goes into some sort of overdrive mode when presented with something it doesn’t know about. The amount of Facebook Shares wouldn’t even be relevant at that point because at least it found something. The only time Shares would even be something worth considering is if there were competing listings for that term.

    So while this is great insight and a cool experiment, we’re back at square one.

  • http://www.twitter.com/saraboargs saraboargs

    great post, thank you!

    I have check on Google Australia and Rand’s original share is actually showing for web search:

    http://www.google.com.au/#sclient=psy&hl=en&source=hp&q=bussangle+nefaribook&aq=f&aqi=&aql=&oq=&pbx=1&bav=on.2,or.r_gc.r_pw.&fp=119d4aa9065bf24b&biw=1440&bih=742

  • http://uk.linkedin.com/in/rashidmamun Mamun Rashid

    Thanks for the article Danny. I just did a search on Google for “bussangle nefaribook” and it seems Rand’s Facebook page, the permalink and his wall are all ranking in normal web search

    http://www.google.co.uk/search?q=bussangle+nefaribook&hl=en&complete=0&pws=0&prmd=ivns&filter=0&biw=1680&bih=937

    I used both chrome (incognito mode) and firefox (safemode); the results are identical at the time of searching.

    Obviously, Google index Facebook fan pages, and I’m sure Google index publicly accessible personal Facebook pages, and may be, just may be use links shared on publicly available personal page as a ranking factor. In this test it is difficult to tell since Rand’s test page is a noindex page.

    To this point, I would like to request that Rand performs 2 more tests,

    1. Remove noindex from the test page to see if it outperform other links for the same search term
    2. Change his wall to only friends view for the test period (I’m sure Rand won’t mind doing so for the greater good of SEO)

    Perhaps this will show us some difference in the kind of personal Facebook pages Google index.

  • http://www.erocket.co.uk Dave Fowler

    Looking at the cached version of the Search Engine Land Facebook page just now, Google can see that you have 19,810 likes for the ‘fan page’ as a whole, and that the most recent share – ‘3 Types Of Facebook Image Ads That Work’ – has 10 likes and 8 comments (some by SEL, some by others) over a 10 hour period.

    It’s a simplistic view, but I would not have thought it overly challenging to capture this type of information to ascertain whether a site has a fan page, how many have ‘liked’ it, how active the page is (frequency of shares and uniqueness of those shares), and what levels of engagement shares achieve (comments and likes they attract), etc.

    How you weigh and incorporate those as quality signals is more challenging, of course, but that is surely in development (if not already live)?

    Drilling down, the ratio of comments to likes could also be a clue as to whether sites are incentivising likes through competitions etc, or whether they are more genuinely given.

    This is such a fascinating area, no wonder it’s got us all debating and speculating so actively. As ever, thanks for cutting through so much of the cr*p and keeping our wilder imaginings (quite probably including mine, above) on track.