Likes vs Links Case Study Results — Someone’s Gonna be Mad.

27 Flares 27 Flares ×

Mythbusters2010 2 Likes vs Links Case Study Results   Someones Gonna be Mad.

In Jan­u­ary, we announced our Like vs Links Case Study to see if the wildly pop­u­lar “Likes are the new links” rumors were true. Appar­ently we weren’t the only ones that wanted to know — over 40 peo­ple sub­scribed to the blog to find out. Finally, the results are in.

And the Win­ner is…


Dis­claimer: By post­ing these find­ings, includ­ing URLs of the sites tested, we hereby end the case study. Any­one can now link to these domains and con­t­a­m­i­nate the find­ings. In case the rank­ings are dif­fer­ent by the time you go and check, keep in mind they’ve been steady for >3 weeks now and we’re over a month out from the time when all the links and likes were created.

seo case study likes links results Likes vs Links Case Study Results   Someones Gonna be Mad.

To our sur­prise, the site that just received links out­ranked both other sites, even the one that got both, Likes and Links.

The site that received 190 Likes and no links at all remained nowhere to be found in the search rank­ings con­firm­ing our hypoth­e­sis that Face­book Likes have no impact on search rankings.

To con­duct this case study, we went on Google trends, like many an inter­net mar­keter before, and found the trend­ing term “Deer Antler Spray.” After fig­ur­ing out that Deer Antler Spray was a health sup­ple­ment and not a clean­ing agent for the prized antlers above your fire­place (yee­haw!), we quickly threw up 3 pages to run our exper­i­ment on. All 3 were semi-exact match domains and were only mon­i­tored for the term “Deer Antler Spray.”

URL Treat­ment Final Rank­ing Received Likes and links >200 Only received links 14 Only received Likes >200

Since we were skep­ti­cal of the effec­tive­ness of Likes, we wanted to give them max­i­mum lead time to show their value. Sites 1 and 3 got their Likes right away.  Sites 1 and 2 got their links 2 weeks later.

To read the whole method­ol­ogy, go here.

Did the Farmer Update Hurt the Case Study,
Or Did the Case Study Hurt the Farmer Update?

Around the time that we promised that we’d pub­lish this case study, Google announced its now infa­mous “Farmer Update.”

We imme­di­ately thought, “Crap. This is totally going to mess up the results,” so we decided to let the exper­i­ment run a bit longer before pub­lish­ing any findings.

Again, to our sur­prise, 2–3 weeks later, noth­ing had changed!

What’s more is that the links we built were from arti­cle direc­to­ries. So if you’re freak­ing out that “Google killed arti­cle mar­ket­ing,” chill. Maybe arti­cle direc­to­ries don’t rank as well as they used to, and per­haps links from them aren’t worth as much as they were before, but arti­cle direc­to­ries still pass link juice and can help you rank.

The proof is in the pudding.

Again, I’m not say­ing that the update had no impact. It obvi­ously did and I know peo­ple who have been affected. With that said, if you’re still run­ning around clench­ing your ass cheeks with both hands, now may be a good time to stop and get back to work.

Flaws of this Case Study and How You Can Help

To be clear, this case study is not the end all be all.

With a sam­ple of only 3 sites, its purely anec­do­tal and nowhere close to sta­tis­ti­cally sig­nif­i­cant. In order to achieve that level of evi­dence, we would need to test many more sites across dif­fer­ent ver­ti­cals with vary­ing lev­els of competition.

One thing I am per­son­ally skep­ti­cal about is that Site 1, which received both Likes and links, didn’t rank as high as Site 2, which only got links. I do not believe that Likes actu­ally penal­ized this page. This dis­crep­ancy can prob­a­bly be accounted for by poor index­ing of links for Site 1.

If any of you sim­ply aren’t con­vinced with our find­ings, we encour­age you to replicate/improve our exper­i­ment and post your own find­ings. Sci­ence gets bet­ter when more peo­ple do it. We want to grow a more sci­en­tific, data-focussed SEO com­mu­nity so we can put the gurus out of busi­ness. In all hon­esty, if you repli­cate our case study and find the exact oppo­site of what we did, we’ll post your find­ings so that the com­mu­nity can eval­u­ate both. Our only agenda is The Truth.

Nonethe­less, we think that this case study is suf­fi­cient to jus­tify a work­ing belief that Likes are not impor­tant for search rank­ings and not a worth­while use of time or resources for a search cam­paign (social is a whole dif­fer­ent animal).

Give Us Your Feedback!

We have a ton more case stud­ies in the pipeline and we’re excited to become your go-to source for data-backed SEO research.

Your opin­ions mean a ton to us so tell us what you think!

Will these find­ing change the way you do SEO?
Are there other more press­ing issues you want to see us testing?

Tell us what you want to test and we’ll do the test­ing for you!

P.S. Don’t hog the goods! Help us spread the word about this case study by post­ing it on Twit­ter and Face­book (use the but­tons on the left).

P.P.S. While you’re at it, don’t for­get to sub­scribe to The Truth blog in the red box below to stay up on future case studies!

11 thoughts on “Likes vs Links Case Study Results — Someone’s Gonna be Mad.

  1. Alex Pyatetsky Post author

    Inter­est­ing. Good find.

    All 3 pages had out­puts from the same spun arti­cle, so “more rel­e­vant con­tent” seems like an unlikely rea­son for the huge dif­fer­ence in rank­ings, imo.

  2. alex

    Another pos­si­ble fac­tor is that google doesn’t want to rank 3 sites on the same ip for the same search term.

  3. Alex Pyatetsky Post author

    Wow, lot of Alex’s on here, haha :)

    Re: not want­ing to rank sim­i­lar terms on 1 server–

    This is pos­si­ble. How­ever, keep in mind that we gave Likes a 2 week head start on links. If Likes would have had an impact on rank­ings, it doesn’t make sense why the site that got links 2 weeks later would have been cho­sen as *the* site to rank on the server.

    Just like dupli­cate con­tent, its usu­ally the orig­i­nal source that ranks.

    Do you rec­om­mend that we split the sites between 3 dif­fer­ent shared hosts next time?

  4. alex

    As the other alex said above, there are too many fac­tors. The main thing to remem­ber is that cor­re­la­tion does not equal cau­sa­tion. The con­tent is dif­fer­ent, LDA analy­sis yields dif­fer­ent results, G def­i­nitely doesn’t want to show mul­ti­ple sites on the same IP for the same phrase unless there’s no other com­pe­ti­tion, G may be putting a sand­box period on exact match style domains since they’ve been get­ting a lot of heat lately, G def­i­nitely doesn’t like 1 page sites, etc. The prob­lem with SEO/link tests is that for a real sci­en­tific exper­i­ment you need to change 1 vari­able while keep­ing every other vari­able con­stant and that’s just not pos­si­ble with SEO since you have to have dif­fer­ent domains and since keep­ing some vari­ables con­stant (like con­tent) would actu­ally ruin the exper­i­ment. So all results have to be taken with a grain of salt and cor­re­la­tion trends recorded over a series of experiments.

  5. Contempt

    I dis­agree com­pletely. On one of my sites I’m still rank­ing #3 after 4 hours (and now 3 weeks later) with noth­ing but FB likes.

    I’m not say­ing your case study is wrong, I’m merely say­ing that there are way too many fac­tors to make a broad state­ment like Links is stronger than links & likes. Com­mon sense should tell you otherwise.

  6. Alex Pyatetsky Post author

    @Contempt: Our claim wasn’t that likes were stronger that links, merely that we had no evi­dence that Likes had any impact on search rankings.

    @Alex: You’re def­i­nitely right. As I men­tioned in the 2nd half of this write up, there def­i­nitely needs to be a series of sim­i­lar exper­i­ments con­ducted in order for there to be any kind of con­clu­sive­ness. Our goal was to test an overly strong claim that has become very pop­u­lar in the indus­try over the past year and see if it had any valid­ity. Some­one read­ing this case study can make their own con­clu­sions from the data, because we’ve been 100% trans­par­ent with the method­ol­ogy, but the whole pur­pose was to put it out there so peo­ple could dis­cuss over some­thing concrete.

  7. Sang Min Park

    Thanks for post­ing this, it’s about time some­one did this type of exper­i­ment. Not really sur­pris­ing that links are still the decid­ing fac­tor; good to know that links from arti­cle direc­to­ries are still worth something.

    About the site with links and likes rank­ing lower; I agree, it’s prob­a­bly just an anom­aly; a higher sam­pling would prob­a­bly elim­i­nate (reduce) such discrepencies.

    By the way, great name Con­tempt; I can see you’re just an insti­ga­tor. Links are stronger than face­book likes, get over it. And it’s not even sur­pris­ing, although I did expect face­book likes to have some impact.


Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>