Back in mid-December, SEOmoz made a number of changes to our use of the rel="nofollow" tag on the internal link structure of this site. We nofollowed dozens of links on many of our template pages to help control the flow of link juice through to our more important pages - the content in the Blog, YOUmoz, Marketplace, and Articles.
SEOmoz's Blog with Nofollow Links Highlighted
As you can see, we used quite a few nofollows to help draw juice away from unnecessary pages. The results?
_
A resounding success! While you can see that search referrals were particularly slow at the end of December and into January (as was overall site traffic), once the holidays passed, the search traffic came rolling in. The bump is approximately 20%, quite a substantial rise from the addition of a tag on some less important links.
While others have certainly reported seeing success with the use of nofollow to control the flow of link juice, I figured this quick example might help serve as a catalyst for anyone who's skeptical about the potential value.
BTW - I do certainly realize that many other factors could have affected increased search referral numbers, and thus this test isn't entirely scientific. However, our search traffic going all the way back to May of 2007 was increasing at an infintessimal rate (Weeks 20, 21, and 22 of 2007 [May-June] had an average of 12,000 search referrals) and this is the only structural change we've made since that time. Thus, don't take this data as gospel truth, but rather as another nudge that nofollow sculpting can have high potential rewards.
p.s. We have done used this same thing on projects for clients with mixed results - several have seen increases of 20-25% in 2-4 weeks, but one of our clients stayed virtually level after implementation.
p.p.s. For more on how sculpting PageRank with nofollow works and why it can be effective, see Si's excellent post on the subject .
Rand, maybe -- but maybe not. I mean, on the one hand you say "a resounding success," while on the other you say there could be other factors.
OK, so you say this is the only structural change you made to drive up search referrals, which weren't building from earlier in the year. You went from apparently 12,000-13000 weekly to around 15,000, and this is the only structural change you've made.
I've had no structural changes and sculpting, and our search traffic is up:
https://searchengineland.com/080102-195756.php
So much for sculpting. Then again, we've been growing in authority from being a brand new site, a difference from SEOmoz. We've also added new content, just as SEOmoz has. And it could be that your new content has added to your search traffic, couldn't it be?
I want much more to convince me. What's different about Week 1 and Week 2. Yes, there's a holiday dip, but you still should be able to isolate which particular content is driving that extra 2 to 3,000 visits from search. That's what I want to see before I buy into the nofollow sculpting.
Of course, sculpting in general bugs me. So I was there with you and the rest when Matt rolled this out as a suggestion people should do, and I thought it was crazy. After all, Google first tells us not to worry so much about doing things for search engines and now we're suppose to tell people to start blocking links? At this late stage? Like Google hasn't ready figured out how to discount a lot of the common links that repeatively show on pages?
Even the concept isn't new. It's just PageRank hording, but using nofollow to do it.
Hey, it could totally be that it IS helping. But it could be other things. That's why I'd really like to see the drill-down on what's making the spike, to better understand if it is down to the scrulpting.
Yeah - I tried to make it very clear in the post that it's only my opinion that this has been the difference maker and that it's very hard to seriously rule out anything else.
One of the other positive elements that I really like is that more of our historical blog content ranks (rather than being in whatever they're calling supplemental now).
As far as sculpting as a practice - it's always been around. I remember a guy from back at SEOChat who swore by using external javascript links to any page where he didn't want PageRank flowing. Nofollow is just making the process a little programmatically easier, right?
Have you seen any change in rankings for different content.
In some industries, we see a huge increase after 26th Dec in search volume and in some, X-mas marks the end of high search volume.
So, have you looked at the keywords driving traffic to your website and checked search volume trend on Google.
Rand,
I'd have a hard time assigning cause and effect here. The main reason why you'd use this to "sculpt" site structure is to improve indexing, but I doubt you'd have any problems with that, since every page probably collects inbound links like mad.
To believe that using nofollow to channel your PageRank a bit can cause a big leap in traffic, you'd have to think that PageRank was a very significant ranking factor, and you'd want to try that on a site that doesn't attract so many incoming links to every post.
If you want to test it here, I'd suggest a concerted effort to write bad posts, lots of posts about your pets, that sort of thing, over an extended period. It won't be easy to get people to ignore you, but it's in the name of science... so come on - take one for the team, dude!
@ Michael Martinez, I really don't know what your problem is, but if you can point to even one (1) false or misleading statement I've made on this, I'd be happy to correct it. OTOH, you've been running around saying that using nofollow is harmful, will cause pages to get dropped from the index, etc - which is somewhere between misleading and just plain false. Look at what Rand just posted - why didn't all of those nofollows destroy SEOMoz?
I love the discussion (although admittedly, while I'll accept the use of nofollow as a tool... I haven't completely agreed with Google's position on it), and I love the idea of looking at metrics and testing...
but I'm not sure I could completely jump on the bandwagon on this one as others have touched on. Rand, I'm glad you pointed out some of the caveats and possible issues in this. In the end, I think it would be hard to argue this as a definitive example either way.
Actually, I think trying to test and measure the impact on a site like SEOmoz is too challenging.... just too many moving parts. In fact, any blog or heavily syndicated or social mediazed site could really be swayed by many factors. But SEOmoz, especially with both the main blog, the YOUmoz section, and the tools, could be subject to a lot of influencers.
I think a more ideal test environment would be a stable site:
Perhaps more informational in nature or even ecommerce, but again, have to watch out for direct or indirect seasonal influencers in either case.Monitoring of both keyword ranking as well as traffic flow, and perhaps even crawling activitiy (do these "sculpted" pages experience increased crawling as they now appear more important?), and monitoring both the sculpted and nofollowed pages.
And then some form of validating based on the "not in place, put in place, remove, put in place" approach. Three months per change? Maybe ideal, but probably not realistic. Could it be 1 month on, 1 month off? Maybe 2 weeks on and 2 off.
Either way, great discussion and hopefully it will prompt additional testing.
That's exactly what I am doing. Running the test on a dynamic, relatively small site. I will share my findings here soon.
I think the shorter periods will work as long as it gets spidered that frequently.
We did it on a large enterprise level site (75k+ pages with no dupes) - nofollowing all of the usual links on the page template (terms, privacy policy, accesibility, etc etc). The results were very similar to Rands. Scuplting the juice flow definitely works, even if you're only saying your terms pages are unimportant!
Rand, you need to track more than overall traffic to prove internal nofollow works. Did number of hits to supplemental results pages increase? Did you see any improvements in ranking? Did your site gain more keywords coverage (ranking for more competitive keywords which eluded you before)?
This post reads like: "I submitted my sitemap xml to Google; two days later, my site was de-indexed" Correlation doesn't mean causation.
When I have an on-site SEO factor I'm trying to get my head around, I often look at how seomoz is coded to pick up any tips that you guys are using on your own site.
I happened to notice that your 'home' link was nofollowed (thanks to Aaron Wall's firefox extension and its nofollow highlights), asked my self "I wonder why that is?", searched "nofollow" in your search box and found this post on exactly that with a very interesting discussion below it! I found the answer to my question (and more!) in a matter of a few seconds. Just thought I'd share this to demonstrate how easy it was to use seomoz. You have a post on pretty much everything!
Anyway... my 2p worth re the discussion about "scientific" claims in seo: it seems to me that nothing in SEO is truly scientific due to the closely guarded nature of the algorithms, and the fact that there are so many factors involved. It's an iterative process of SEOs getting together and discussing what has worked for them and gradually coming to a consensus (thats why this site works so well). If people like Rand refrained from exploratory posts like this, waiting for some "scientific proof" a year down the line, then nothing would ever get worked out...
Seems like this is igniting some interesting discussion. I like your points Danny and Michael. I love the point where you say "At first the Search Engines don't want us to design the sites for the Search Engines" and now they want us the webmasters to tell the Search Engines. It's definitely too late to do that. Depending on the structure of your website (no matter your website is small or big) it could be a very small task or a close to impossible task to tag the not so wanted pages as nofollow.
But I am convinced to your point Danny. Is the situation so out of control that the search engines now want the webmasters to tag their pages "for the search engines".
Michael, your point is very interesting to prove the test. The question is who would be willing to do this test over those many months, pressing the on and off switch so many times. Is it a simple on and off switch in SEOMoz's CMS ? I don;t think so. On a sidenote, I am thinking of running this test on one of my sites which has not been touched in over an year, and just do this and watch what happens, and then remove nofollow and then watch again and then add again and see what happens. Might prove the point, because that website does not have the Authority status as SEOMoz, no content being added, no links being added (not that I know of) etc.
All in all, great discussion.
Nice post Rand - it's really interesting to see this in effect. Also - congrats on the lift in traffic :-)
What was the breakdown like for the other search engines? Did you only see a lift in Google or was it across all of them? Did any search engine not respond to the nofollowing?
Tom - I used Google because Yahoo! and Live haven't fully crawled much of our archives, and thus still record link juice as flowing in the old patterns. I'd expect that I can probably judge them better in another 40-60 days.
BTW - This is the same pattern we saw with our clients' sites. Google's new super-speedy crawl means we usually see architectural benefits like this from them before anyone else.
That makes sense. Once the inner pages of the site are indexed again using the new nofollowed links it would be interesting to see a comparitive shift across the different engines.
Tom - had to go back and check the stats for all engines. Here you go:
So, Google clearly the leader, but some benefit from all the engines apparently.
Thanks for sharing, Rand.
Here is my experience with nofollowing links on a huge site:
1/ within 6 weeks we've been experiencing steady growth;
2/ within next 6 weeks we've been experiencing steady decline.
Now we have just a bit higher traffic than 3 months before (nothing except onsite optimization has been done during those 3 months).
Hey Rand, how come you didn't notice anything when Google changed things around "more than a year ago"?
Painful question, I know, but someone had to ask. Sorry.
I agree. Seems like this Matt's poston Pagerank sculpting clarifies the situation. If you have nofollow links a part of pagerank is than wasted.
Really useful - thanks Rand. It would be interesting to dive into the data and see what you are ranking better for than previously (but that would involve a heck of a lot of rankings reports!).
That's a good point - would be interesting to see if this just had a positive effect on the inner pages or whether there was an increase seen for the homepage and main pages as well.. Still, might require a fair amount of analysis! (and the rankings for the main pages is harder to analyse in a vacuum since there are plenty of other factors which could influence that..)
using no-follow tags surely helps in terms of controlling outbound links - to make a site more competitive (can't think of any word to say it right). =)
ExSEOllent post Rand!
After all the bad press the nofollow attribute has been receiving as of late this comes as a validation of it's purpose, a no entry sign for bots that can be used to steer traffic through the meat of your site.
Some would question the validity of your observations gauged on the data you used, but the proof is in eating the pudding and your stats reflect that there was a marked increase in traffic.
Observing and analysing this over the long term is going to be interesting. This debate is far from closed!
It is difficult to tell if something makes a difference, or if it is a combination of other factors. Something that might be interesting to see that might reflect on the effectiveness of sculpting with nofollow is your crawl rates.
What frequency of crawls are you seeing on your 'important' pages vs your nofollowed pages, and how does compare to before the sculpting?
Because we know that page rank impacts crawl rate this might be a helpful metric to see what effect the sculpting had. If the sculpting has worked, we would expect to see an upswing of crawling on the 'important' pages and a downswing of crawling on the nofollowed pages.
I know this is an old post but was directed to it from a QA thingo. It is seriously good for my site as I have god-knows-how-many links and most of them are unimportant (and not helping rankings).
Thanks
have seem the same kind of positive results more time than not.
Rand,
Controlling the out bound links with no follow tags, i hope its really working, but not for designs , I know, since I started learning all these things now. After the process, now how is your sites status, improved a lot ? and how much percent we can keep the do follow links in our site any guess from you ?
No problem Rand, Danny actually clarified that for me as well...just wanted to make sure! Keep up the good work.
I've read a few posts where Rand and Stephen Spencer both recommend nofollowing pages that don't require a lot of link juice to rank e.g. about, privacy, terms,etc. However, i see that SEOmoz and Netconepts don't employ nofollows on their own privacy, about, contact us pages...
Hmm... We used to - must have been a code change that didn't get run past the SEOs! Thanks for the heads up :-)
Having read all of the comments I'm also going to try some sculpting. Everything I've learned so far from SEOmoz has proved to be beneficial. I figure it can't hurt ...
Page Sculpting does prove that Page Rank is not dead.
And Thomas, a lot of people have lost a lot of money thinking that Google "should" be able to do something. As a matter of fact, a lot of SEO experts make claims based on what Google "should" be able to do...lol. But the point is, they have their limitations and SEO in some cases is about how to understand those limitations and then acting accordingly.
As it relates to this topic, it means using the no follow tag to help Google decide where to assign Page Rank where the content will be of most use to users. Obviously, there are not a lot of organic searches for your privacy policy (unless you did something wrong..lol), your shopping cart page, your verisign icon if you have one, etc. So using the tag to devaluate the importance of those pages based on the known Page Rank algorithm, is good SEO IMO.
I believe its a great strategy, and something I've been trying to implement into our corporate site since I read about it here a few months ago. Thanks for sharing the results Rand!
And yes there are a lot of factors too look into here, as Danny points out above. But in all reality, sometimes the analytics can't be perfectly explained and you just have to go with your gut feeling. Besides who knows your site better than you?
BJ
Definitely a great idea and good to see the numbers. Seems like I will have to spend some time analyzing what pages you decided to nofollow and which ones not, and then figure out the "why" part. If you did it, there was a reason and if you did not do it, there was a reason.
Thanks for spiking the idea of using the nofollow this way.
I have been working on it too. Using nofollow tags for not important pages on sites. So far I have seen a little increase in Google traffic, non in Yahoo or MSN... Could it be something like Google rewards you for using a nofollow tag and sees sites using it as a good sign of a trusted site? Thanks Rand for the post.
I'm curious why you nofollowed the "View by Author" links. Aren't those perfectly legitimate paths to archival content that you might want to give out some link-juice to?
I think the point is that those links don't add anything anchor text wise so passing link juice through those links is not that great. The preferred route to the blog posts is via the category pages which DO provide relevant anchor text and those pages actually want to rank for something.
I think that's the case anyway. Only had a quick look at it - might be way off! :-)
If there's one thing I hate, it's when a Critchlow is probably right :)
It probably would've helped if I scrolled down a bit more and looked at the original screen. I see your point: the category links are going to hit the same content in a more targeted way (keyword-wise). Makes sense.
Great point Tom_C, I have a question. When you have a link to the homepage, with "Home" as anchor text, is it better to put a nonfollow there for that same reason?
Thanks maria :-)
In theory I'd say yes, but you have to be careful when nofollowing links back to the homepage since you don't want to create orphan pages or reduce the power of your homepage (in most cases - see below). I think andy beard made a post about crippling your site using nofollow which is well worth reading (I'll see if I can dig it out)
In fact seomoz does this (nofollowing home links that is) so they at least think it's worth doing but I think (and this is pure theory/guesswork) that this is only worth doing since they have so much content that resides on the blog pages (and other inner pages) which they want to promote. In some ways, the home page of seomoz isn't as important as for a lot of other sites so they don't mind reducing it's weight.
I could be wrong about that last bit though and would appreciate it if Rand or someone could swing by to confirm that :-)
edit: this is the andy beard article about nofollow and dangling pages but having re-read it I'm not sure it's as relevant as I thought it was (though still essential reading) https://andybeard.eu/2007/11/seo-linking-gotchas-even-the-pros-make.html
Tom thanks for the mention of the article.
Is it relevant?
Actually yes, because many SEO experts advise blocking off duplicate content pages, and unimportant pages with robots.txt or meta noindex.
Those pages will still accumulate juice, though how much juice depends on how Google are attributing juice to dangling pages before they are taken out of the iterative calculations.
SEOmoz structure changes?
The big problem is that there were quite major changes done not just changes with nofollows to less important content.
In the past on every page there were category links, some relevant, some irrelevant.
Those are no longer present, and categories are used to channel juice down to posts from the front page.
The problem is that the category archives are not very flat, so juice only reaches the most recent articles in each category. I am sure search traffic is probably up on more recent articles, and reduced a little on older content.
SEOmoz also isn't a good example, because of the high number of deep links that the domain receives, quite often deep links to articles. Many of those articles are now receiving juice only from outside sources, and some of those outside sources are also not highly optimized.
In addition, because of the reduced number of internal links on each permalink page, more juice is flowing out via links to external sites.
I think the biggest benefit now would be to provide flatter archives, and to slightly increase the number of internal links that leave ech permalink page.
"Single" pages such as a disclaimer or contact page can be used to help channel juice, as they don't have to have external links, and can focus juice on specific targets, or to cycle juice back up to the home page.
This post will probably give people more ideas on internal linking structure
As to comments about whether this all works, it is hard to tell now Google have messed up the reporting from /* so that it is inconsistant from one domain to the next. I worte about that before Christmas comparing indexing of various SEO blog.
I think the most important factor in all of this is ranking better on terms that convert. It doesn't matter if there is an increase or decrease in traffic. The most important is making pages rank that bring in conversions of one kind or another.
By funneling juice around, it is possible to make specific landing pages rank better.
Edit:added anchor text becuase the URL was being cut off
Hi Andy - thanks for stopping by and explaining a few things. I didn't mean that your article wasn't relevant to the discussion (it's highly relevant wherever you're talking about nofollowing pages and site architecture!) but instead I didn't think it was actually that helpful in the specifc case of nofollowing links back to the homepage of the site...
I'd be interested to hear your thoughts on the pros and cons of nofollowing some links back to the homepage?
I do it on niche sites, and recommended it to a reader who I will eventually write a case study on if results ever stabilize enough to really tell whether there was a significant improvement.
I also have it on my blog currently, though I really should change the way I do my header to be a full graphic logo and use CSS - ranking for part of a word or a name doesn't make much sense, and it can make snippets look ugly.
A home link with useless anchor text such as home or click here doesn't really help, so if you can add another link, but with useful anchor text somewhere else on the page, such as in the footer or navigation, that helps, and then you can stick nofollow on the one with poor anchor text.
That being said, I have never experimented with multiple links on the same page to the same destination URL, and for instance on my blog I have had a situation where among the comments I have had 20 or 30 links to my home page with the anchor text "Bndy Beard"
Does Google really count all those links? I have no idea.
I do know that pages can rank purely from internal linking and on page optimization.
Just one example on my blog is for the single word monetization. A few days ago my tag page was on the front page of results in Google from where I am, currently it is on the 2nd page. As far as I know it has never had any links from external sources, other than scrapers :) , though I did use that tag fairly recently.
It is not a tag I use that frequently, so it is a reasonable example of the benefit of internal linking with good anchor text.
If it can help a single page, the benefit for a home page should be much higher.
Nice commentary Andy.
Are you sure this isn't due to the uptick that tons of sites saw right after the new year because of how google is now handling supplemental results? You could probably check this by comparing the number of search terms that visitors are finding your site on with the number around 2 months ago for example. You might find a lot more strange long tail results rather than better placement for the same results. It's also my experience that there's always an uptick in traffic in Jan.
We made the same changes back in October or so, but saw the same traffic changes in January. According to the theory, that should otherwise have been November.
coolness. Haven't thought of using nofollows like that.
Nice work!
Does this work better for large sites (i.e. getting deeper pages indexed), or would you see similar benefits on smaller sites (e.g. increased ranks for the homepage)?
I'm going to guess it works better on larger sites, as most of the benefits we see aren't in the head of the search demand curve, but rather the "long tail." However, I have to say that I haven't worked on a very small site in a long time, so it's tough to say - certainly worth a try, IMO.
This is very interesting data, Rand - thanks for sharing it.
I was curious myself to see how much value could be pulled from controlling the flow of PageRank with the nofollow attribute. I have to say that my initial estimate was "not much," but I still read-up on it and discussed it with some people to ensure I had a good grip on it. This is one of those cases where I'm happy to find I'm wrong - controlling the flow of PageRank with the nofollow attribute is absolutely a valid optimization technique.
"One of the other positive elements that I really like is that more of our historical blog content ranks (rather than being in whatever they're calling supplemental now)."
That, to me, is the strongest signal that sculpting page rank with the nofollow tag works.
I'm still very skeptical about all of this, but knowing it doesn't hurt rankings makes it worth testing out in my opinion.
Good to see the case study on this.
Rand, I could be wrong but if I remember in a post awhile back, perhaps it was even a Whiteboard Friday, where you had mentioned sculpting PageRank wasn't recommended in all cases and could even backfire. Do you still feel this way?
Yeah - I think so. If you're new to it, don't understand the principles or get mixed up and make an error, you can end up killing your long tail traffic by focusing too much juice on the most important pages and forgetting to flow it into your archives or UGC areas (at least, those are the most common errors I see).
Great post Rand!I was wondering what effect this would have on a small site?My site is a week old, has only 4 pages, and receives 70 unique visits a day through organic searces. It has no trouble ranking due to a low amount of competition, but I want to add a new section which will have a large amount of competition and will have trouble getting external links to it. Should I use nofollows on the page that ranks well to every page apart from the high-competition page?Thanks!H (I'm a newbie to SEO!)
Nofollow attributes work considerably better on sites with thousands of page and sites that have authority on longer tail keyphrase focused pages.
Example:
abc | def | ghi | abcdef | defghi | abcghi
funnelling links from abc, abcdef, and abcghi to higher category pages a, b, and c will garner you a more authoritive page and reduce the confusion that Googlebot would have had on the site had you not nofollowed to create a better 'funnel'.
If you only have a few pages you can start funnelling the impact is going to be less.
Did I make any sense? Rand, do you agree or disagree?
Brent
I too have been using internal nofollows for about the last 6 months, and it seems to be helping, but as you mention there are so many factors we are trying to improve on at one time it does make it hard.
I posted last week (https://www.seowrench.com/nofollowed-internal-backlinks-in-webmaster-tools/) about how they are counting even the internal nofollowed links as backlinks in Webmaster tools. Has anyone else noticed this?
Jim
Someone posted somewhere that the backlinks listed in Google WMC are all links not just links that pass link juice. So you'd see nofollowed links and regular links.
Brent D. Payne
nofollow has the universal appeal of putting a choice in linking, directly into the webmasters hands. this of course, is not absolute, it's "a" choice and not "the" choice, there are no definites yet, but it is good to see GG moving in a direction of inclusion, even if we're not completely sure to what degree.
I pitch this concept to everyone I talk to. It is very powerful and should be used as a 'big rock'.
It is easy to see how link juice flows through your website (with no-follow, robots.txt etc.) using e.g. A1 Website Analyzer. However the effect on search engines... Mmmm. I tend to think it could be problematic to no-follow all links to e.g. privacy policy...? And shouldn't Google be able to filter out common sitewide links like that? But worth testing, especially if you are suffering from e.g. "terms" page coming up before your main/product page in search results :)