Over the past few years, the topics I’ve researched, written, and spoken about have evolved. One of the most common questions I’m still asked about is the relationship between social media sharing and SEO performance. Thanks to Rand and his awesome team at SEOMoz, I got access to their Mozscape API and was able to actually start to answer these questions in a scientific way.
To complete this analysis, I compiled a database of more than 25,000 URLs that had been shared at least once on the three major social networks (Facebook, Twitter and LIinkedIn), were at least a month old, and had at least one incoming link.
First, I looked at the relationship between the number of times a URL was Tweeted and the number of incoming links it had pointing to it. I found a convincing positive relationship. Those URLs that got more Twitter love, also got more link love.
Secondly, I looked at Facebook and found, somewhat unsurprisingly, almost exactly the same effect. Facebook popularity is related to inbound link popularity for URLs.
Finally, I looked at LinkedIn sharing. Of course the numbers are much smaller here due to sharing activity being much more common on Twitter and Facebook, but I still found another positive relationship.
For all of the “big three” social media networks, I found that social sharing had a positive relationship to incoming links pointing to a URL. This result is basically what I expected to see. However, when I took a step back and compared the actual Pearson’s Correlation Coefficient of the sharing on the three networks to inbound links, what I found was surprising.
While all three networks did have a positive correlation, the strength of the relationship was strongest for LinkedIn. So, while LinkedIn may be the least obvious choice for sharing activity, it is still incredibly important for marketers also interested in SEO performance.
Looking for more insights into online marketing? Don't miss the free webinar on August 20th with Rand from SEOmoz and Dharmesh from Hubspot: The State of SEO and Internet Marketing in 2012.
Wonder how google + would have tested out?
Also,
When adding a biy.ly or how.ly URL shortener does this affect the link? or does the link still give the site the same value?
I also had these same questions after reading this post. But you asked them first. Thumbs up for you!! :)
Same question to be asked ? , Thanks
"When determining which URL shortener to use remember that you want a shortener that will do a 301 redirect from the short URL to yours. That way you can keep as much of that link juice flowing to your own site as possible. Also, be sure to use one that gives you some analytics about clicks and such, like bit.ly."
Quote from https://www.seomoz.org/blog/the-social-media-marketers-seo-checklist
and this article digs deeper into the different shorners and 301's
https://searchengineland.com/analysis-which-url-shortening-service-should-you-use-17204
Oh sweet thx :)
Kyle your shared links are just fabulous. Might be you won't believe but your sharing had changed my vision for social media. Still now i was only believed that socials are just the opportunity of being social to the customers or with the rest of the audience. But we also get link love from them is really something new to me, Well this is the best part of the MOZ we learn something new each day. Cheers.
And that SE Land article is really a treasure.
Thanks Again.
Yes this was the case in the past but I would guess that google would have seen this as a potential way to manipulate and started passing even less value through bit.ly links. They would only have to look at the domain before removing the value. Also it has been fairly well documented that redirecting between domains carry even less weight due to people purchasing domains and 301ing.
What would stop you creating a type of bit.ly, then when it gets loads of links you then do not 301 to that link but to other domains that you own which are on the same topic to get the value for your own domains?
I am sure Matt has elaborated on it in Past. These urls are 301 redirects and pass the link value if links are not nofollow!
You are right! Matt Cutts did emphasise (in video q&a) the importance of ensuring that any shortening service implements a 301 redirect: https://blog.web-media.co.uk/social-media/google-short-urls-are-fine-for-seo/ As pointed out by Seot0p below some don't.
Yes, such shortened links give the site the same value if returns 301 server response (some shortening services may cheat and uses 302 redirect, which doesn't pass the link juice).
@SEODinosaur -- Google+ doesn't have enough activity to correlate their data (joke). But seriously, here is some data from Marketing Pilgrim:
For every 100m users, the avg. number likely to share a story:
Twitter: 197.3
Facebook: 41.8
LinkedIN: 15.2 (and I thought LinkedIN was terribly low)
Google+: 6
https://www.marketingpilgrim.com/2012/08/call-out-the-urban-spelunkers-google-is-officially-a-ghost-town-infographic.html
Sweet this is very helpful data.. Only thing to left and ponder about is how biased is Google when it comes to using their social network versus others...
Big mistake using link shorteners unless you host your own https://wpsites.net/best-plugins/how-to-setup-your-own-url-shortener-for-wordpress/
I am finding this statistical analysis troublesome on several levels.
First this type of analysis requires the use of a null hypothesis. Which I do not see here. You cannot use r to simply prove an assumption correct. You must be testing again something, not inductively proofing.
Next, Pearson's r must meet the following criteria (below) and the data points are typically shown in scatter graphs around a line of best fit. This is not what is shown here. Here is a standard plot graph.
Here I am not seeing that #4 or #5 were accounted for also find showing the line of best fit without data unusual (outliers). Controlling for outliers in a Pearson's r is very important and if not controlled for other statistical measures would be more accurate.
Finally, using bar graphs to show a person's coefficient is unusual given the need to show data around a line of best fit.
In addition, if comparing the three networks you would most likely need to jump to a linear regression analysis were the variables are held at a constant to control for spurious relationships. At the least you must convert your r for comparison
And lastly I see no data on the tests for statistical significance here which if not run means you cannot be sure what the Pearsons r means. If a .5, but the statistical sig tests come back inconclusive you have no relationship.
You also do not show the r2 which tells you the coefficient of determination which gives you the ability to say yes a relationship is indeed causal. It is not sufficient to just compare r.
NOTE I am not saying you did not account for these, but would like to see how they were accounted for and if it is here and I missed it I apologize in advance.
Thank you!
Pearson's r REFERENCE
There are four assumptions that are made with respect to Pearson's correlation:
(see our Types of Variable guide for further details).
(see our Testing for Normality guide for further details).
(jump to this section here).
(jump to this section here).
From https://statistics.laerd.com/statistical-guides/pearson-correlation-coefficient-statistical-guide-2.php
Can you establish cause-and-effect? No, the Pearson correlation cannot determine a cause-and-effect relationship. It can only establish the strength of the association between two variables. As stated earlier, it does not even distinguish between independent and dependent variables.
How do I report the output of a Pearson product-moment correlation? You need to state that you used the Pearson product-moment correlation and report the value of the correlation coefficient, r, as well as the degrees of freedom (df). You should express the result as follows:
where the degrees of freedom (df) is the number of data points minus 2 (N - 2). If you have not tested the significance of the correlation then leave that section out of the results.
Can I determine whether the association is statistically significant? Yes, the easy way to do this is through a statistical programme, such as SPSS. We provide a guide on how to do this, which you can find here. You need to be careful how you interpret the statistical significance of a correlation. If your correlation coefficient has been determined to be statistically significant this does not mean that you have a strong association. It simply tests the null hypothesis that there is no relationship. By rejecting the null hypothesis you accept the alternative hypothesis that states that there is a relationship but with no information about the strength of the relationship or its importance.
What is the Coefficient of Determination? The coefficient of determination, r2, is the square of the Pearson correlation coefficient r (i.e. r2). So, for example, a Pearson correlation coefficient of 0.6 would result in a coefficient of determination of 0.62, which is 0.36. Therefore, r2 = 0.36. The coefficient of determination, with respect to correlation, is the proportion of the variance that is shared by both variables. It gives a measure of the amount of variation that can be explained by the model (the correlation is the model). It is sometimes expressed as a percentage (for example, 36% instead of 0.36) when we discuss the proportion of variance explained by the correlation. However, we must never write r2 = 36%, or any other percentage. We must always write it as a proportion, e.g. r2 = 0.36.
It's great to have the data behind this but it just seems like commonsense doesn't it.
The Social Echo is a reality. Good things happen when your content is shared and seen by more people. You know that a certain percentage of those impressions will result in links - and the good kind too.
The difference in platforms is interesting. Perhaps LinkedIn users are more professional in nature and have a greater propensity for having their own blog? Or maybe it's a sample bias. And of course it all really depends on the audience you're targeting.
Ensuring that your content is easily shared and optimized for social platforms is vital. It's not just SEO, it's good marketing.
Great thoughts, AJ. I have to agree that it's vital that we also focus on creating great content that people want to share, but also make it super easy to share. I'm big into WordPress blogging (for myself and my clients) and have found that while there are dozens of social media plugins available to make post sharing possible on all the major social networks, it's worth the extra time to find the right plugins that support the most integrated experience possible between the blog reader's social sites and the blog posts they want to share.
For example, the Twitter API will pop up a dialog box that includes the title of the post and the (full, not shortened) URL. After the user clicks to share, the dialog box returns another message asking the user if s/he wants to "Follow" the blog author (if s/he is not already doing so). This extra functionality helps to grow the author's Twitter following.
Thanks for your insights.
LinkedIn is so overlooked. It seems people are in love with Twitter and Facebook but we've had good results with articles getting shared on linkedIn and it brings decent traffic depending on the topic.
Definitely topic dependent, business/leadership topics and their sub-topics seem to be the only things that get any real traction on LinkedIn, given it's professional audience.
Also, somewhat relevant xkcd ;) -- https://xkcd.com/1034/
Hehe, I love it when xkcd is relevant to real life. :)
I have also under estimated Linkedin. I don't have Linkedin account yet. I am just playing with Twitter's tweets & re-tweets, FB shares & likes and Google+ posts and its +1. I have read Karl pearson correlation chapter in school time at that time I don't know its that much useful.
I'd love to hear more about how to get maximum benefit from sharing on LinkedIn. I'm new to the service.
It looks good in a graph, but it really depends on what links are measured. You could turn it on its head and say that URLs with a lot of links tend to get shared more on social media. Which comes first, the chicken or the egg?
@steve, my thoughts exactly. It would be interesting if this test were done on relatively new links that didn't have any inbound external links to start with.
What came first? The link or the social share? It would be nice to see social vs. links over time.
It would be interesting to dig deeper into how social is leading to the creation of links or vice versa. It would be reasonable to assume that social sharing helps people to discover great content which then leads to links, but it would be nice to see how this happens and over what timescales (and of course how this helps the page in question rank...)
Agree, a timeline which shows the evolution of links/social would give greater insights.
Interesting research, but I can only infer the following from it: popular websites normally have more inbound links and appear in social media more often. That is no new insight. Furthermore I would expect popular sites to receive more page impressions, vistits etc. - but does this require a detailed survey?
Couldn't agree more.
Correlation isn't causation.
Only valid conclusion (IMHO) is: the more popular site is, more links it gets.
Linkedin vs the rest part is interesting, but I would repeat the experiment in a few months with a larger sample.
BTW, this post was written by Dan Zarrella of HubSpot. :-)
This can be sooo argued... Depends on who the sharers are, who they are followed by, how many followers, the nature of the niche (i.e. how active it is on the web and how likely people in it are to have any web properties they are able to place a link on), etc. Too many unknown variables to come to any solid conclusions. And if you do the analysis cross-niche it's like comparing apples to oranges.
Hi Dan!
Interesting data you show here. I was wondering about two things, which you maybe could clarify about.
You say that you have a database of more than 25K URL.
1. Are these URLs mixed with a widespread on-page topic?
2. And are these URLs in different languages?
Thanks.
I was wondering the same thing. It's easy to say we grabbed 25k URLs, but are they from various industries?
We all know Tech/SEO/Marketing sites/blogs/pages get way more links due to a lot of their target audience being people who own websites.
I'd be interested in knowing if this study covered a wide range of topics or "a lot of the same"
Hi,
Why you did not include gogole+ ? We all know how much it is important as social sharing. This comparison convincing positive relationship, so now it is clear in our mind that social sharing also provides an inbound links and click.
It's an interesting article but I'm not 100% convinced by your argument.
I do a lot of feed syndication via Feedburner - I tend to link the accounts to Twitter - which is also linked to Facebook profiles.
Whilst this helps with news / guide / blog content sharing I've found that the main source of in-bound links tend to come from websites that syndicate the news feeds or bloggers / writers who are subscribed to them and pick out useful content for their own blogs / sites - not necessarily social networks.
Having said that I think social networks are very useful if you want to pick-up more subscribers - so the circle is relatively obvious (as long as you target correctly). If you want publishers to pick-up your content then social has an obvious impact - but it's also about how you retain those subscribers.
Nice analysis, one thing I notice that you didn't talk about though...
Even though there is a stronger correlation with LinkedIn, the quantity of shares to links is a lot lower (hidden by the difference in scale on your graphs). There appears to be an almost 1:1 relationship between shares and links for Twitter, not much less for FB, but a 4:1 relationship for LinkedIn. So while the correlation is not as strong, it appears that Twitter/FB is more effective (assuming there is a causal relationship here!).
Also, a couple questions on methodology:
1. Did you include the links from the Social Networks in your inbound link count? If so then you're double-counting and of course you'll get a close correlation
2. Did your LinkedIn shares include LinkedIn Answers and Groups, or just status updates?
Good question 1.
Twitter recently dropped linkedin partnership , so it could be useful to know if your data are before/after Linkedin statement
Nice stats, but is this really relevant since HubSpot has a lot of followers? Will it do any good for web properties that are weaker than HubSpot?
Isn't it just as likely that many of the links being shared in these so-called social media channels are being promoted? And that therefore links are correlated with shares because those URLs are being marketed and subject to other promotional efforts, not that the sharing is organically causing links?
For example: I don't know about you all, but in my observation, normal internet users generally don't share links on LinkedIn; marketers do. So wouldn't it be obvious, then, that LinkedIn would have the highest correlation with links? It has the highest concentration of sharers who are probably marketers.
How many of these shared URLs are not being socially shared, but promoted as part of marketing efforts that include direct link-building?
I absolutely agree with AdriaK. It might be that those two measures (LinkedIn shares and # of incoming links) coincide, so the correlation isn't based on a causality.
Can you perhaps give us Chi-square and p for these correlations (to see if the finding is significant); what about r-square? Have you considered performing a permutation test for r and a bootstrap? Personally, I feel just looking at r might not give you the complete picture. But I'm sure you didn't just look at it - however you mostly presented and argued with that. So maybe you could give us the other statistical measures to give a more convincing case.
Other than that: great work! Thanks for the share.
Edit: I just saw that a few of my points have been raised by standardgirl already. I agree with her post as well. On the r-presentation I'd say that might be a bit fuzzy - sure it is unusual (mostly due to the fact that bar graphs make it seem like a continous measure, which it isn't) yet I don't think it's that bad, really. Again, I'm sure you did a thorough statistical investigation but only tried to make it easily digestable for the casual reader. Then again, it might help to provide these other measures as well to be more convincing.
As per usual, correlation isn't causation - which I think Doug alludes to with his first comment. However, it does highlight that LinkedIn probably has greater scope to improve our SEO efforts than we sometimes give it credit for.
I'd say don't go out now and ignore FB/Twitter for links, but ensure that LI is kept as part of your social mix, something that requires a bit more effort now you can't tweet directly to your LinkedIn feed.
One thing I wonder about is if each share is measured the same. If someone is highly active in say Facebook, and has 1000+ friends does that share carry a little more weight then someone with 0 friends? No two inbound links are exactly identical in strength so does this hold true for social media?
What about Pinterest? Yes, I agree on Linkedin. Its very powerful for our markets and customers, less so on twitter and facebook.
This kind of seems like common sense, but good job setting it up to be easily understandable. Good to see linkedin in there as well.
does anyone knows a great guide to using Linkedln for SEO and getting traffic/clients?
I think I biggest factor in LinkedIn is making sure your profile is setup properly. Here is a great article on that.
https://www.seomoz.org/ugc/optimize-your-linkedin-profile-for-best-results-howto
I would look into Lewis Howes for anything related to LinkedIn
Great article there ... But why was GOOGLE+ left out?
As mentined, should have good leverage ...
Ah, great analysis danzarrella; is it for high authority sites like hobspot or any general websites, bcz here I'm trying lot's ways and not getting that much of inbound as i expected. Any ways thanks for sharing. oh yes you didn't show here G+??
I am beginning to see some very positive results using social media and especially Linkedin. If a website is struggling to rank for keywords then optimising your Linkedin profile can be a short cut to getting some attention on Google.
Really interesting research Danzarrella, this is something new. What I think is Topic is what matters more while it comes to sharing & inbound links through social networks.
I don't agree with the stats above. No doubt link shorteners are translated - anyone can see that in Google WT. Agree that Google did state that it's watching Social Shares - 2009 anyone? - but that doesn't mean it still does. I've seen more backtracking than forward movement, e.g. dropping twitter and FB shares in SERPs. And thats pretty pointless - just because there's 900 million people on fb doesn't mean we're all now connected. And the potential to do that is ridiculous.
Most social media platforms (by SM, don't we largely mean Facebook, twitter, LinkedIn?) are either completely blocked (e.g. FB, Branch) or completely nofollowed (twitter, FB) and also diverted via things like l.php or app.php (e.g. FB and LinkedIn).
I don't see a lot of Facebook inbound links in my GWT, I don't see blog posts or pages that are shared frequently on twitter rank better. The data is fudgy.
But here's the real concern: This will only open the flood gates to spam. Social Media - and why I remain so sceptical about its use for SEO is that its far easier to game. If we thought SEO spam was bad - you ain't seen nothing yet. 180 million fake accounts on twitter? Followers aren't important? The first suggestion that Google was measuring (measuring doesn't mean "has impacted yet") was to see how important a link was from "influential" people. Klout anyone?
The second worry: The more we tell people "You should do x, y, z" - the more we'll create a plasticky carbon-copy web - and by default, cause Google to change.
Content Marketing take note: the problem isn't content (the user decides) - the problem is getting it in front of people. Trying to hit a certain segment by writing entertaining content for their friends will drop conversion rates, just as blogging broadly does.
Great post! I just started using Linkedin.
Nice article, and thanks for the "share" (pun intended).
I've just yet to see any hard data from anyone to REALLY distinguish correlation/relationship vs causation.
From Will Critchlow: https://www.seomoz.org/blog/do-improved-social-signals-cause-improved-rankings
I always had been thinking that twitter is the top social sharing platform but after reading this , i was really amazed that linkedin is topping up the charts as well.
I am sure that lots of people her , have overlooked the linkedin a number of times.
Very interesting findings. It points to one very important fact though and that is social media has a lot of power and links from social media is just as important as inbound links.
Wouldn't imagine LinkedIn would be the winner, as FB naturally contains more followers.
My LinkedIn friends never flow to my website via my shares, while Fb takes the lead.
Oh well, can't argue with statistics.
Maybe I just need more circles and followers... but LinkedIn is the hardest place to get them!
This is a very good post.
I've found that the more targeted Twitter followers you have and tweet to, the more quality links you get.
I use a tool which finds quality followers by scanning their profiles for common interests.
This tool has found me over 40,000 followers which i have written about in more detail recently on my blog.
No point in getting followers if they're not interested in what you are writing.
I know you can do something like this with Linkedin but Facebook isn't as easy.
If this study was only for HubSpot, then certainly the conclusions about the value of the different social media outlets can only be for HubSpot's niche.
HubSpot may have content that's great to share on LinkedIn (mostly because LinkedIn is a way to network, share and gain professional knowlege, hence HubSpot's webinars).
However, if your cupcake business is on LinkedIn, how likely are your blogs, photos and other content going to be shared and give more links?
I'd like to see more about the methodology or the study replicated for websites of various topics/industries and if the data still supports the conclusions made here.
LinkedIn is great for making connections with other professionals. Conversions can be high depending on your field since its not a hang out like other social networks.
Agreed. I personally seem to get a decent number of clicks/likes/comments on LinkedIn shares when I'm sharing my own content on SEO (i.e. new blogs posts). It definitely depends on the industry though, and whether it's B2B or B2C (the former performing much better in my experience).
I noticed that now most Online SEO analyses include some metric for social media. Maybe because people like to insert url in their status message, still I don't see the benefit if the nofollow tag is set.
I manage a car service in sf and they rarely use their social media and still they always show up in searches related to limousine in SF
Thanks for researching this Dan, like you mentioned in your post this has been a constant discussion in our community and this is something for all of us to look over and learn from. I am not surprised by the performance of LinkedIn at all. We represent a b2b brand that is getting a ton of traffic from LinkedIn. As you all know as a marketer you must follow the 4 P's of marketing. For the brand I am speaking of, we knew they would get the most traffic out of Linked (the people AND the place) and we were right (I love it when that happens) it has generated a lot of shares and traffic. I am looking forward to seeing what other statistics you compile from this research.
That is surprising about LinkedIn. Looking forward to the webinar!
Great post! I just started using Linkedin. It feels like it is not the place to overpost, but if done correctly, it obviously can be a great tool. Thanks again for this post. I will be showing this graph to some of my clients who refuse to use Linkedin!
Very cool... Do you think the sample set size of linkedin influenced the positive correlation? For example, in baseball - if a batter only has 10 at bats and gets 4 singles, his batting average would be a whopping .400. MLB sets a minimum for the fewest amount of at bats to be considered a league leader. Make sense? Great article to read though!
Aren't all LinkedIn shares routed through a LinkedIn URL?
nice!