PageRank does leak and if you have indexing problems you may want to keep your eyes on inbound/outbound ratios. It wouldn’t be high on my list of todos though - ROI would depend on the type of site you’re working on.

External links to authoritative pages does nothing to increase your authority. The only reason you might link to another authoritative page is to improve user experience.

Nofollowing outbounds to prevent leaking is not something I’d recommend. Nofollow if you’re selling links or if you’re referencing a page but not endorsing it. Nofollowing external links may not help you preserve PageRank under the new rules.

That said, one way to prevent PageRank leak is to rewrite links in javascript, so that people with js will see links while others without js (including SEs) will not. Not a zero risk tactic but should not be high risk since you’re not displaying different content based on user agent or IP.

Links to flash player download pages for me is borderline. Macromedia probably will never link back to me and they already have plenty of “votes” so I’m not really robbing them of anything by nofollowing links that point to their site. It depends on what type of site you’re dealing with - number of links, (real) PageRanks of the pages those links are on, the total number of links on each of those pages - but my guess is you’ll only see a marginal improvement if any.

Another thought/question related to this. In reading earlier notes on the Google Algo, it said since pagerank is in effect “leaked” when you do off site links. It said then that external links should be kept to a minimum (I have software that measures these ratios for instance and now wonder if that’s even a necessary excercise). It also said that putting these links from as few pages as possible was preferred. Conversely, I have read that external links to authoritative sources increase your authority via theme relevancy. In what situations do you think each of these are true? Does containing outbound links to fewer pages help?

And given the above, even without retaining page rank by using no follow (if the current page rank/no follow changes are true), wouldn’t it be wise to no follow external links to avoid “leaking” your existing page rank? Would the above situation still justify the effort of no follow under the “new rules” if they turn out to be?

For example, We’re redoing some websites right now where we have many links for users to download say a flash player and was trying to decide whether to no follow them or not.

Just wanted to leave you a note here as well…

I figured out how I copied PageRankBot results. The time it worked when I tried it, it was with a smaller site, on larger sites the Windows Clipboard can’t handle copying and pasting all the data. So here’s what you do:

Select a partial number of lines. (1000? Dunno the limits of clipboard memory)
Press Ctrl+C
Paste into notepad with Ctrl+V
Select from Notepad, copy, and paste into Excel.
Repeat.

Like I said, with a small site you can do this all in one copy/paste step, but with a larger site you need to space it out. There’s probably an easier way to do this (some clipboard program or bumping up the default memory for Ctrl C), but the good news is it works, and you can save the results into a spreadsheet!

Hi Gabriel,

If you read Matt Cutts’ recent interview with Eric Enge, you’ll learn, as I did, that robots.txt and META noindex are pretty much useless when “sculpting” PageRank. So you have a few choices: don’t link to “untargeted” pages, like TOS/about me/privacy policy - not something I’d recommend because they’re useful to visitors. Those pages usually hoard 90% of a site’s PageRank. That means less PageRank for your article pages. Too little PageRank flowing downstream means supplemental results.

This isn’t about “hoarding” PageRank by not linking out or nofollowing every outbound or counting the number of links on a TBPR 6 directory page. It’s about channeling PageRank to article pages to lower the chance of them turning supplemental.

Second option, like I said, is to link to those fluff pages for usability sake but nofollow the links so you can save your PageRank for some other page that’s more important to you.

“Reminds me of newbies talking about pagerank hoarding, except you’re obviously not a newb and what you’re saying is clearly sophisticated.”

As Matt Cutts also explained recently, if your site is a TBPR 2 site (home page = TBPR 2), this tactic isn’t going to have much impact, because you don’t have much PageRank to play with. But if your site is a TBPR 7 site with thousands of pages, controlling PageRank flow can have a dramatic impact. How is that?

As Dan Thies recently explained, to rank high, you hunt for anchor text and a decent amount of authority (PageRank). You can hunt for targeted anchor text outside of your domain, which can be difficult via viral campaign and costly via paid link campaign and ineffective via recip link spam (if you don’t know when to say when). Instead, you gain authority (more link pop) via viral marketing, without concerning yourself with anchor text, use that authority to get thousands of pages indexed, and then use those pages to gain thousands of targeted anchor text. Think about some of the reasons why Wikipedia ranks so well.

My article explains how you can improve the depth of index penetration by using nofollow. Neither robots.txt/META noindex can help you do that because they don’t block PageRank. The goal here isn’t to hoard PageRank; you can’t create more PageRank by adding nofollow. You are just controlling where the PageRank goes and how much falls on each page.

Like they say, it’s all in the wrist, not in your head :) You got my email.

Thanks,

Aaron

Agreed. But I’m not talking about blocking a ton of internal links here; just 4~5 category links that, on a blog, I find pretty useless to begin with.

“For a typical small business Web site, there are no unimportant pages.”

Michael, that happens to be true for one site I’m being paid to work on, but Dan made a good point that affiliate marketing sites usually don’t bother nailing top positions with their category pages. I rather rank on the first page with 10,000 product pages than with 10-20 category pages. On my blog, in particular, there are some old posts I wrote that aren’t valuable to me at all. I don’t mind shoving them into the supplemental index.

“They may get a few pages crawled more often but they could easily achieve that without resorting to nofollow.”

I don’t need every single page linking indiscriminately to a navigational URL (e.g. home page). True, you can avoid that scenario without using nofollow (e.g. Andy Beard’s Sandcastles). Still, when you have 1,000 sitewide links to the home page, adding a nofollow to 2-3 of them isn’t going to kill your site either.

The more internal linkage, the better. The more internal links you have pointing to pages, the sooner and more often they are crawled and indexed.

“I’ve never seen Google miss a crawlable page on any of my domains.”

I have.

However, you and Dan have both missed a major point in all this. PageRank may be driving the crawling and the indexing (two different actions), but a crawlable link has to be found before the page on the other end is even queued for crawling. And that is just as important after the page has been indexed as before it is indexed.

When Google goes through a major data dump as it just did in August, the more infernal links you have the less stress your site undergoes in the dumping process. Also, the more internal links you have the sooner you get your pages recrawled and reindexed.

For a typical small business Web site, there are no unimportant pages. Search engine optimization is not about favoring one page over another. It’s about optimizing every page for search.

People who follow this bad advice about using nofollow on their own pages will not realize any SEO benefit. They may get a few pages crawled more often but they could easily achieve that without resorting to nofollow.

Michael, good point about nofollow turning off link anchor text. That’s a pitfall. If a website’s second-tier pages are ranking well for keywords in the internal anchor text, sitewide nofollow will obliterate those rankings. That’s why I proposed a partial nofollow. Also, if I manage to gain deeper index penetration from this tactic, that’s more anchor text than if I had only a few of my pages in the main index.

What to you is good architectural design?

“that means the chances of an important page being crawled and indexed are reduced if you use “rel=’nofollow’”

I’ve never seen Google miss a crawlable page on any of my domains. In fact, the click-path from the home page to a leaf node is shortened when you cut out the middle-man, so I’d expect leaf nodes to be crawled more often. If PageRank is the main determinant of crawl depth/frequency, then sending more PageRank to the leaf nodes might encourage Googlebot to crawl them more often. Also bots can’t tell a third-tier page from an uncle page. They’re all links on Googlebot’s crawl queue waiting to be crawled. Lastly, the only time I expect bots have trouble discovering URLs is if there’s only one path to a URL and that URL is the only one linking to another URL, and so on, in a continuous chain.

This prevents crawlers from following those sibling links on the second tier, hence reducing the frequency of crawl for those more important pages (as well as decreasing their chances of being indexed).

“2. Use nofollow on links on third-level pages to second-level pages so that a third-level page passes PageRank to its parent page but not to any other pages in the second-tier.”

This prevents crawlers from following links to those aunt and uncle page, thus reducing the frequency of crawl for those more important pages (as well as decreasing their chances of being indexed).

“3. Tiered Pairing: To prevent second-level pages from losing too much PageRank, you can link them in pairs: e.g. page A with B, C with D, and so on.”

This reduces the number of followed links on the second tier, thus reducing the frequency of crawl for those more important pages (as well as decreasing their chances of being indexed).

“4. Circular Navigation: To circulate more PageRank on the leaf level, link them up in circular faction, so page A links to B and C, B links to C and D, etc.”

That has always worked better than just having leaf pages link back to their parents. Why? Because it increases the number of inbound links to each page.

NOFOLLOW is not about PageRank, it’s about pretending a link doesn’t exist. Matt Cutts says that Google won’t follow the link if it’s NOFOLLOWed, and that means the chances of an important page being crawled and indexed are reduce if you use “rel=’nofollow’”.

This kind of bad architectural design is not a substitute for strong internal linkage. People need to understand that when they turn on NOFOLLOW they turn off crawling and link anchor text, not just PageRank.

This all circles back to John Andrews’ recent post (with an “seo-whiners” slug): “when Google talks, listen up. Google is practically telling you how to rank.”

Here’s Eric Enge’s take.

Googlers had already been saying it was lack of pagerank, which could be the result of duplicate content but not necessarily.

Since alot of SEOs still look cross-eyed at nofollow and PageRank, I doubt either of them will catch on at the grass-roots level just yet. Too many SEO-related ideas are irrational and emotional, not tactical. I’m glad to see Rand Fiskin coming around though.

“I’d guess probably 1/4 actually indexed?”

Even less than that. His second-tier pages are TBPR 4s, as is his home page. He’s got a relatively decent link profile; he only links out editorially, and he’s got a handful of Yahoo! Directory links. Many of his landing pages rank on the front page of Google. He just needs a few more pages in the main index.

As long as Google considers PageRank important, it will be important to understand and use that knowledge to get more of your important content into the index. I have students “discover” that they actually belonged on the top of SERPs after we did some basic internal linking changes - with no real link building campaign at all.

Your client is a good example - 2000 pages, and I’d guess probably 1/4 actually indexed? You can get that whole site indexed with the right structure and a relatively small amount of incoming links.

“Nofollow is now, officially, a “tool” that power users and webmasters should be employing on their sites as a way to control the flow of link juice and point it in the very best directions.”

Interesting times.

“With only 129 pages (I didn’t spider, just Googled site: so that may be way off the mark)”

Right. Originally I had around 200 pages but I seeing what happens if I trim down. The site I’m being paid to work on involves around 2000 pages.

“One of the problems with analyzing things internally only, is that it actually does matter where PageRank comes in, and how much.”

I agree. With my tool you have an option to compensate for that by simulating backlinks to a specific page, but its far from accurate. Honestly, I’m not exactly sure how to go about simulating that. It makes a huge difference where the entry points are. For example, with this domain I only have a few backlinks to my root URL; most of my backlinks either point to my blog home page or to individual blog posts. A commercial site might have 99% of its backlinks pointing at the home page. Like you said, things like that make a huge difference as to how PageRank flows internally.

I’ll have to go back in and make sure that my description of the 3rd-level-push idea is completely accurate in the book. I’m going to be really mad at myself if I screwed that up…

With only 129 pages (I didn’t spider, just Googled site: so that may be way off the mark) your blog isn’t much of a candidate for any of Leslie’s schemes (you may even do better SEO-wise with 100% cross-linking), but it’s fun to play with this stuff anyway.

One of the problems with analyzing things internally only, is that it actually does matter where PageRank comes in, and how much. The “one way valve” strategy (home page has one link to a site map which links into the rest of the site) looks OK if you only do internal analysis, because you added a page, so there’s more total PageRank.

When you actually do the math on what happens after you push PageRank (external inbound links) into that single-link home page, then you realize that you’d have been better off pushing PageRank directly into the sitemap instead of losing some at the “home page,” because it doesn’t all make it through the one-way valve. Which basically means you might as well make your home page look like a home page.

Yeah, I do. I’m actually more worried about my client’s site (which is commercial with very few outbounds), but I’ll have to come up with a solution to that.

Even if you don’t factor in any discounting of sidebar links, you really need another 10 or even 20 internal links on this page.

You have 4 external links, I would aim to have 20 internal

Sandcastles seems complicated, but it can be achieved with just a few plugins and fairly simple theme changes.

I have no idea how Google handle pages with massive amounts of links . My page on dofollow plugins is just crazy now with 200 external links, and a huge growing tag cloud at the bottom.

Right, I overlooked dealing with the category pages.

“With no incoming or outgoing link you are concentrating PR in your date archives.”

All my archive pages (save two) were robots.txt disallowed though. Now I updated robots.txt to disallow them all, So PageRank should be “trapped” at the third-level. I used allow/regexp in my robots.txt which I think PageRankBot choked on. Since there are a couple of backlinks to those archive pages, I might remove the disallows and nofollow them instead.

If you look at the third screenshot (which now reflects reality more closely), you’ll notice there’s no PageRank concentration. The URLs you see there (except the last URL - the SEO category page, and the domain home page) are single post pages. The tactic I used was linking semi-randomly at the third-level, and completely eliminating sitewide links.

One problem I created, like you said, is not having enough internal links per page. Ideally, I want to have no sitewides but have as many internal links as possible per page (up to 100-150).

Another problem, like I mentioned, is lack of focus. But if I use a random function to randonmly remove nofollows on top post links, I can “staccato” my most important pages. It might sound a little extreme, but the beauty of that is it gives me more control over PageRank distribution while I avoid compromising user-experience.

I noticed Sebastian’s post as well, which made me kinda wish I had powerpoint. I gotta admit I do enjoy reading about “under-the-hood” stuff.

It took me a while to decypher the PageRank calculator grid. If I understand it correctly, with the Sandcastles approach, you have only the tag pages linking back up to the home page, category pages don’t crosslink with each other, and none of the content pages link back up to the category pages. The home page gets some weighted attention due to alot of links linking into the tag pages and the tag pages linking up to the home page, but the set up is much more controlled than having every page on the domain linking back up to the home page. It makes the default Wordpress look like a broken template.

I actually started looking into this after suggesting third-level push to a client having supplemental problems, to make sure I wasn’t steering him wrong, and to have a post he (or his programmer) can read which explained internal link structure in more detail. I ended up learning alot in the process.

I don’t think you are really disagreeing with Leslie

At the category and date archive level you are creating a Campbell Mini-ball

Then you are channelling down to single pages from both, and then channelling back up through the data archives.

With no incoming or outgoing link you are concentrating PR in your date archives.

We now introduce some links coming in, and some external links on each page.

So lets be fair and add in some more pages as dummies
The additional 9 pages do not have incoming or outgoing links just like my tag pages, so offer a similar (though not exact) benefit.

You will see that you don’t have quite as much pagerank concentrated in your single pages, and far too much in your date archives.

Here is a version of Sandcastles to compare, but this one I have removed the sitemap so it has an identical number of pages.

On the plus side, if you reversed the role of the category and date archives, you would end up with some fairly powerful categories and a lot more relevance being passed.

I know this is all theory

One of the advantages of doing everything with template changes and plugins is you can have multiple themes and switch between them to then use your tool.

I am going to have to get your tool up and running now, plus finish an article linking through.

I need to test out what Sebastian is doing as it wasn’t very clear - I really enjoy this kind of research.

Not really Andy. It took me a couple minutes to hack that (though it took a bit longer to write the solution out in this post). I’ll check out the plugin though; thanks for mentioning it.

“also still have a ball linking structure in your 2nd tier.”

Ok…according to Leslie Rohde’s “Dynamic Linking”, a ball linking structure is basically a structure where all pages link to each other:

“the sub-pages will … therefore not need as high a PR value, nor as much Link Reputation. It’s our home page that needs the most links, but our linking strategy has placed it on a par with our internal pages.”

You’re right. Links on my category pages aren’t nofollowed, but still the impact is minimal because now I only have the home page linking into those pages.

I disagree with Leslie though; when dealing with supplemental results, the sub-pages need as much PageRank as you can afford to give them.

“One gotcha for many people is that they will try to combine this with blocking off their archive pages from indexing, thus reducing the number of internal links per page.”

Good point. That can lead to more PageRank bleed.

“Did you see my Sandcastles linking structure?” Yes. I can’t give you my take on it till I re-read your post a couple times though :)

I tested my tool’s Java version using several SEO blogs, including yours I think. Unlike the PHP version, it reads robots.txt, META noindex tags, and handles 301 redirects. Please let me know if the antisocial plugin nofollows are still not being picked up.

It is much better to use the add_link_attribute plugin for sidebar elements

One gotcha for many people is that they will try to combine this with blocking off their archive pages from indexing, thus reducing the number of internal links per page.
The end result is they will create a great sacrificial linking structure.

Did you see my Sandcastles linking structure?

I need to get your tools working and finish up my optimization (and get the final theme published), when I tried your PHP script the other day it barfed after 2 hours, so I will give the javascript version a shot.

I did notice that for some reason the php script was not assigning nofollow to the links from my antisocial plugin, where the image links to the social bookmark sites are nofollowed.