Leveraging Pr On Site Options

[mountainbound]

Was wondering if anyone had suggestions on how to put to good use the decent page rank some of my pages have. My home page has PR7 (but I can't take credit for that), and a few other pages on the site have PR6 & 5s.
Is there any way to funnel what I see as a nice natural resource back into my site to increase other pages as well?
Thanks,
MB

[SearchRank]

First of all, don't place too much emphasis on PageRank values.

With that said, just make sure you have a good navigational structure. In addition well planned out site map, breadcrumb trails and hyperlinks in copy where it makes sense to add them should all help increase internal link popularity.

[mountainbound]

Thanks SearchRank. I did create a map a few months ago, but not sure of the SEO effect it has had, though from a usability standpoint I'm sure some users are thankful. I registered with google sitemaps today and am getting it indexed at the moment. Hope this was a good move.
Maybe I'm still a little fixated on PR, but I realize that there is a weighting algorithm that enables a page to extend some of its popularity to every page that it links to, but those pages all have to share a slice of the PR pie.
Do you know if general "top nav" links, or footer links, detract from the overall PR weight that a page has to give?

[Michael Martinez]

Maybe I'm still a little fixated on PR, but I realize that there is a weighting algorithm that enables a page to extend some of its popularity to every page that it links to, but those pages all have to share a slice of the PR pie.

They do all have to share the pie, but what many people get confused about is how the pie is sliced out.

The PageRank valuations in all their messy forms are only intended to be approximations of a probability distribution.

There is no way to know the true distribution. Google and anyone who estimates PageRank takes a best guess at it. A lot of people mistake the changing values that are produced by successive iterations as a shift in actual PageRank, but it's not.

That misunderstanding has confused some people into believing that if you start linking out too much, you lose PageRank. Outbound links only confer PageRank. They don't diminish. But all PageRank scores are constantly decreasing anyway because the number of Web pages is constantly increasing.

The sum total of all PageRank is 1. As we create more Web pages, that sum total is stretched more and more thinly across a vast sea of documents.

You cannot hoard PageRank, and you cannot dilute your own PageRank as fast as it will be diluted by all the pages everyone else is creating.

Every second, thousands of new Web pages appear on the Internet. Every minute, tens of thousands of new Web pages appear on the Internet. Every hour, hundreds of thousands of new Web pages appear on the Internet. Every day, millions of new Web pages appear on the Internet.

It's pointless for people to obsess over how they are linking out from their pages.

[Ron Carnell]

Outbound links only confer PageRank. They don't diminish.

Sigh. That's patently untrue.

You cannot confer a finite resource (must add up to one, remember?) without taking it from somewhere else.

The real myth is that diminishing PageRank is necessarily a bad thing. It's not, because PR is a lot like money. You have to spend it to make it. No one ever built a successful business by putting all their money in the bank.

But all PageRank scores are constantly decreasing anyway because the number of Web pages is constantly increasing.

The addition of new pages to the Internet pool is largely irrelevant and rarely that simple.

PageRank is a snapshot in time, insuring it is always finite because you can't go back and add new pages to that snapshot. It's also relativistic in that each score only has meaning within the context of all other scores within that particular snapshot. A thousand bucks in 1960 isn't the same as a thousand bucks in 2006, but fortunately the two never have to compete against each other. Whether someone tells you they have a thousand dollars or they say they have a PR5, the numbers lack meaning outside the context of a specific snapshot in time.

As pages are added to the Internet, the pool isn't simply "stretched more and more thinly," but rather is simultaneously shifted to the right, much in the way inflation shifted the value of our thousand dollars from 1960 to 2006. Why? Because virtually none of those one million pages added today are going to be PR9 pages by tomorrow. It's just one more variant of Grehan's Rich get richer scenario. When you add a lot of brand new pages to the pool, the relativistic shift will be to the right rather than to the left. Put another way, an individual number will decrease slightly, but the relativistic value of how much it is worth increases more than slightly. Yesterday, your page was higher than 800 million other pages. Today, even though your PR might appear marginally lower, your page is higher than 801 million pages. It's all relative.

"Leveraging PR" is about trying to put it where it will do you the most good. For most sites, in my opinion, that's a complete waste of time.

While we all know it's a gross over-simplification, let's assume for a moment that internal PR is a function of internal links. The more internal pages that link to Page N, the higher will be the PR of Page N. Since most of the internal pages probably link to the home page, we can guess the home page will have the highest PR. Since only a handful of pages link to big-blue-widgets-in-denver.html, we can assume it will have a low PR.

Relevancy is always, always, always going to trump PageRank.

The pages that are most relevant for most key phrases are likely going to be buried several layers into a site hierarchy. They won't have much PR and, because they are highly relevant, they don't really need PR to rank well. If someone searches for "big blue widgets in denver," they're going to find the highly relevant internal page, not the much higher PR home page. Relevancy trumps PR. I think just about every professional SEO will tell you THIS is where they concentrate their efforts. It's the meat and potatoes of SEO and PageRank is almost completely irrelevant in this arena.

Most site architecture is designed in a hierarchy, from the home page, to general categories, to more specific categories, down to individual and very specific pages. Not coincidentally, this distributes PR throughout the site in directly inverse proportion to relevance. If you completely ignore PageRank and create an architecture that works for your visitors, you're going to end up with low PR on highly relevant pages and higher PR on more generic pages. That's exactly what you want.

Alternatively, if every page links to every other page on the site, PageRank is going to be more evenly spread throughout the site, sometimes to the extent where every page has essentially the same PR. Using a site map, especially if every page links to the site map, also tends to spread PR more evenly. There aren't many advantages to spreading PR evenly, but for most sites it really won't matter one way or the other.

In my opinion, the ONLY time it makes sense to try to manipulate internal PR is if you want a page to rank for a more generic key phrase. It's easy to be highly relevant for "big blue widgets in denver," but very, very difficult to be relevant for "widgets." In fact, it's impossible and the only time PR is going to trump relevance is when relevance isn't really possible. If you want your home page to rank well for "widgets" it's going to take a LOT of links to do it, which will obviously be reflected in your PR. Internal links won't ever be enough to do that, but neither should they be ignored in highly a competitive contest.

If you've designed your site architecture for visitors, you probably already have every page on the site linking to the home page (and if you know SEO, the anchor text is going to use the right keyword). By completely ignoring PR, you've already done it right.

That's not to say, however, that you can't do more. For example, if every page on your site links to your Privacy Page and your Privacy Page links only back to the home page, you've essentially doubled the internal links back to the home page (minus a significant dampening factor) and increased its PR by some indefinable amount. The flow of PR throughout a site can be manipulated, though never without some cost (your category pages go down more than your home page goes up in this example), and it's almost never worth the effort. Successfully ranking for "widgets" probably isn't possible unless you really do have one of the best widget sites, internal links not withstanding, and even if you do eventually rank well for it you'll probably just get more untargeted traffic rather than more conversions.

While it takes money to make money, that doesn’t mean we can afford to spend money foolishly. Similarly, I think there are almost always better ways to spend your time and resources than chasing PageRank. At the end of the day, relevancy always, always, always trumps PageRank.

This post has been edited by Ron Carnell: Mar 9 2006, 05:26 AM

[Michael Martinez]

Sigh. That's patently untrue.

You cannot confer a finite resource (must add up to one, remember?) without taking it from somewhere else.

The algorithim does not work that way.

When Page A's links are used to calculate Page B's PageRank, Page A's PageRank is not altered or diminished in the least.

The mistaken notion that PageRank changes because of outbound links is pervasive but quite contrary to the literature.

[mountainbound]

In my opinion, the ONLY time it makes sense to try to manipulate internal PR is if you want a page to rank for a more generic key phrase. 

Ron

This is the scenario I'm facing. I have a very broad industry keyword that my site does not rank on at all. So I'd like to channel some of my page rank to the page that deals with that broad area. I appreciate your sentiments on relevancy. In this case, I have a very relevant page, but it seems like it could use some page rank help to get it on the map.

Wanted to ask again, do footer and header links absorb page rank? ...if so then every page on my site is being bled by at least 15 links off the bat.

[GBRD]

You cannot confer a finite resource (must add up to one, remember?) without taking it from somewhere else.

The real myth is that diminishing PageRank is necessarily a bad thing. It's not, because PR is a lot like money. You have to spend it to make it. No one ever built a successful business by putting all their money in the bank.

Ron is absolutely correct and commerce is not a bad analogy.

To expand upon this idea it is worthy to note that the act of passing forward PageRank when considered in an isolated context gives the appearance that PageRank is in fact being lost, however when we examine the effect across the entire link graph after multiple iterations it can clearly be seen that this is not the case.

If we look at the individual PageRanks across the link graph as they approach convergence many seemingly impossible scores will appear, this however is also the result of isolating a subgraph from the larger context. The larger context of the entire link graph must be considered as a whole to allow for proper interpretation.

In other words its completely pointless to argue the effect of applying the PageRank formula to isolated page scenarios.

(IMG:http://www.highrankings.com/forum/style_emoticons/default/ph34r.gif)

[Ron Carnell]

This is the scenario I'm facing. I have a very broad industry keyword that my site does not rank on at all.

If it doesn't rank at all now, I'm sorry, but it won't rank any better by manipulating internal PageRank. In a nose-to-nose horse race even small things, like PR, can sometimes mean the difference between win, place or show. If your horse is still trapped in the starting gate, however, smacking it with a riding crop is a complete waste of time.

I think you're first going to need a lot of external inbound links using your generic keyword as the anchor text to get your mare into the race. When you get it showing on page two of the SERPs, you "might" be able to push it to page one by funneling more PR where needed.

It's all very iffy, at best, and in all but the most unusual circumstances, competing for generic keywords is a lot more effort than it's worth. Only you, of course, can make that final call.

Good luck, though. (IMG:http://www.highrankings.com/forum/style_emoticons/default/smile.gif)

This post has been edited by Ron Carnell: Mar 13 2006, 05:01 PM

[Michael Martinez]

Ron is absolutely correct and commerce is not a bad analogy. 

Ron is not correct for the same reason that so many other SEOs are wrong on this issue.

PageRank is not in any way diminished or altered when it's calculated from one set of links to another. The algorithm is a unidirectional adjustment. That means, what changes only changes on one side of the transition.

No PR flows from a page to another page.

The classic PageRank algorithm is actually quite simple:

CODE

2.1.1 Description of PageRank Calculation
Academic citation literature has been applied to the web, largely by counting citations or backlinks to a given page. This gives some approximation of a page's importance or quality. PageRank extends this idea by not counting links from all pages equally, and by normalizing by the number of links on a page. PageRank is defined as follows:
We assume page A has pages T1...Tn which point to it (i.e., are citations). The parameter d is a damping factor which can be set between 0 and 1. We usually set d to 0.85. There are more details about d in the next section. Also C(A) is defined as the number of links going out of page A. The PageRank of a page A is given as follows:
PR(A) = (1-d) + d (PR(T1)/C(T1) + ... + PR(Tn)/C(Tn))

Note that the PageRanks form a probability distribution over web pages, so the sum of all web pages' PageRanks will be one.

PageRank or PR(A) can be calculated using a simple iterative algorithm, and corresponds to the principal eigenvector of the normalized link matrix of the web. Also, a PageRank for 26 million web pages can be computed in a few hours on a medium size workstation. There are many other details which are beyond the scope of this paper.

PageRank is only an estimation of the probability that a surfer will randomly arrive at any given page by clicking on links without purpose. The estimate is based on the number of documents in the collection and the number of links in the collection.

When the calculations are done, the technical papers indicate that all pages start out with the same base PageRank value. With 10,000,000,000 documents you divide 1 by 10,000,000,000 and every document starts out with that extremely small number.

You run through the database and take out all documents that have no outbound links. They will be treated at the end of the processing as if they link to all other documents.

You run through the database and take out all the links that go to documents which are not in the collection. These "dangling links", as they are called, may actually point to real documents that have not been included in the collection, or they may point to non-existing documents.

The dangling links can be factored into the process or ignored. Either way, you'll get variances in your estimation from actual PageRank. Also, just because a document doesn't exist doesn't mean someone won't try to visit the document by clicking on the link. So the assignment of PageRank to unindexed documents would actually provide for a much more accurate estimation than merely treating the dangling links as if they don't exist.

If dangling links are to be included, then they, too, would be processed last and their unindexed documents would be treated the same as all the documents which have no outbound links: they would be handled as if they link to all the other documents in the collection.

At this point, you run multiple iterations of the PageRank calculation across the remaining documents and links. Each iteration moves the estimated values closer to the actual values in a pre-Calculus-like sequence of calculations. There is as yet no way to determine PageRank through an Integral (that I have read about -- I have no idea of what they may actually have figured out by now).

The crux of this "PageRank is passed from one document to another" argument is grounded in the complete misunderstanding that PageRank can be altered or adjusted by someone changing a document's linkage. That is simply not the case. No matter how many links you put on your document, its PageRank is going to be determined strictly by the links pointing to it.

In the classic SEO example, you are given a closed system with 4-10 pages where someone cleverly takes a link from page A and changes it so that PageRank passes around the closed system.

In the real world, this doesn't happen. There are peaks and valleys, spikes and low points, call them what you will. Whole sections of the Web are unrelated to other sections of the Web and you cannot get from those sections to each other simply by clicking on links. It doesn't happen.

Of course, the classic PageRank algorithm incorporates the "damping factor", which assumes that at some point the random surfer gets tired of clicking on links and stops, and may start again at some unspecified point of origin.

When you add a link to your page, your PageRank does not go down. It doesn't change.

Given a collection of X documents and L links, you can change the PageRank of all documents in the collection from one calculation period to another by either changing the number of links or changing the number of documents or both. If we assume that all that changes between today and tomorrow is that 1 link is randomly added to a document somewhere in the collection, that the link is unique to that document and it points to another document in the collection, the target document will realize a boost in PageRank but the linking document will not realize any decline in PageRank, except as all other documents' PageRank is diminished by the presence of the additional link.

However, comparing 2 sets of documents with different links in no way validates or justifies the incorrect claim that a page loses PageRank through its outbound links. In the calculations, the sum total of all PageRank cannot exceed 1 and yes it is a finite value being spread very thinly across a large but finite collecition of documents. The redistribution of PageRank occurs throughout the collection regardless of where the links are placed. All PageRank values are adjusted.

But as the total PageRank never increases, the average starting PageRank can only decrease over time as the indexed document collection grows. That means that overall, on average, all documents experience a decrease in PageRank, including the link-rich documents that accrue new links at a faster rate than other documents.

The only way a document can increase its PageRank is to accrue links faster than the collection grows in both documents and links. There are, in fact, many docments which do accrue more PageRank -- we see this happen all the time as new content becomes popular and people link to it. Those documents are receiving PageRank from all over the place, not just the documents that link to them. But they will eventually hit a plateau point where they are no longer accruing new links faster than the collection is growing and their overall PageRank will, each time the PageRanks are recalculated, begin to decline gradually as the collection continues to grow at a (now) faster rate than their inbound links grow.


Now, that's the base theory. It no longer applies to reality. Google has acknowledged through Matt Cutts that "link-selling sites can lose their ability to give reputation (e.g. PageRank and anchortext)."

If that's the case, then clearly PageRank isn't being calculated according to the classic algorithm any longer (and its relevance to actual probabilities is questionable).

Google also delists sites. Presumably, excluding them from the index means that links pointing to their documents will be treated as dangling links.

Google also penalizes sites. We don't have enough information to know how links from penalized sites are handled. I would hazard a guess that links from penalized sites are not counted.


People need to stop worrying about whether adding links to their pages or not linking out is going to affect their PageRank. Your PageRank is affected by the millions of new documents and links that are added to the collection every time Google updates its PageRank database, and your PageRank is affected by all the delistings and penalties that are applied and/or lifted every time Google updates its PageRank database.

Ron and I could go back and forth forever on these minutiae, but they are truly not important regardless of whom you want to believe. The point is that most of what affects your PageRank is out of your control.

[GBRD]

 
Ron is not correct for the same reason that so many other SEOs are wrong on this issue. 
 
PageRank is not in any way diminished or altered when it's calculated from one set of links to another.  The algorithm is a unidirectional adjustment.  That means, what changes only changes on one side of the transition. 

 

I decided to actually perform a re-calculation (in sub-graph visible mode) across several communities in our simulation index while reading this statement for the second time (for dramatic effect.)

and it looks like...
wait a minute...

Oh, sorry actually it appears that the nodes are both dynamic and inter-dependant. And what's this...

Oh, it looks like we have adjustments occurring bi-directionally across multiple vectors as well. I'm sorry. I think you're confusing the PageRank formula and the iterative algorithm that is used to apply the formula.

If you had decided to actually read my post in context (s-l-o-w-l-y), you would have ascertained that I was not discussing the literal passing of PageRank, but rather I was explaining why that illusion is created when you take calculations out of the larger context.


(IMG:http://www.highrankings.com/forum/style_emoticons/default/ph34r.gif)

[Michael Martinez]

If you had decided to actually read my post in context (s-l-o-w-l-y), you would have ascertained that I was not discussing the literal passing of PageRank, but rather I was explaining why that illusion is created when you take calculations out of the larger context.

Interjecting a non sequitur into a lengthy discussion, starting it off with a very brief statement of agreement like, "Ron is correct", invites that kind of response.

Ron: "you cannot confer a finite resource without taking it from somewhere else"

You: "Ron is absolutely correct"

Doesn't work, since you took Ron's statement out of context:

Michael: "Outbound links only confer PageRank. They don't diminish."

Ron: "Sigh. That's patently untrue.

"You cannot confer a finite resource (must add up to one, remember?) without taking it from somewhere else"

I'll stay with the original context, thank you.


Now, let me sum it up briefly in one definitive sentence. It doesn't matter who agrees with it or not, as the literature speaks for itself:

A page cannot not confer PageRank.

They all confer it, none of them lose anything by conferring it, all they are doing is "voting" the full amount of their earned PageRank through their literal or assumed links.


In real application in today's Google, of course, that is not always the case. Google is blocking PageRank from pages that apparently continue to earn it.

Basically, if you want to hoard PageRank, sell links and get caught. That is the only way to do it.

Until Google figures out that people are gaming the system for that reason.

[Ron Carnell]

... the literature speaks for itself:

It does, indeed, Michael. You just aren't listening to it.

Take 100 pennies and spread them around to ten people. Person A then gives two pennies to Person B. You want to explain to me, please, why you think the sum of everyone's money should now suddenly be 102 pennies?

Similarly, the sum of PR for all pages in the index cannot be 1.0005. You can't create pennies out of thin air, nor can you create PageRank out of nothing. It is a finite quantity, determined solely by the number of pages in the index.

The key to understanding why a page's PR can increase while still maintaining a sum total no greater than unity is in GBRD's last post and is directly derived from the literature you are so quick to tout. You shouldn't really have to be able to wrap your head around the results of iteration, though, to understand that 100 pennies is, and forever will be, a finite quantity.

[Michael Martinez]

It does, indeed, Michael. You just aren't listening to it.

Well, obviously I'm paying good enough attention to get the facts straight.

But I'll welcome any citations you care to post, with links back to the sources that anyone can follow.

In the meantime, I've provided sufficient explanations and citations here and in other discussions to show that I do read the literature -- the reputable stuff, not the SEO-written papers that fall apart at the first bad assertion.

[Ron Carnell]

Well, obviously I'm paying good enough attention to get the facts straight.

Straight, Michael? You're reading and parroting the literature in the same way someone might read Tolkien and then proclaim to the world it's just a child's simple fairy tale. You're entirely missing any sense of depth, blindly waiting for Peter Jackson to spoon-feed you his own interpretations so you can avoid doing any work. You're not really reading Tolkien, you're just skimming it to get to the good parts. You want citations and links back to the sources that anyone can follow? Why? Because it's easier than thinking?

Imagine a page with three links, all pointing to other pages in the same site. How much of that page's PR is conferred to each of those links? It's in the literature, Michael, and implied by the formula. Do you need a paragraph number? Okay, cool, ignoring for a moment the unknown damping factor, we can all agree then that each of the three links will confer one-third of that page's PR to the target pages. See, that didn't take much real thinking.

What happens now when you add a fourth, out-bound link to the page? I'm afraid that's not spelled out in the literature, Michael, so you'll have to do some original thinking. There are no citations or links I can offer.

Will that added out-bound link also confer one-third of the page's PR on its target page?

Of course not. That would take us up to four-thirds and be like coming up with 102 pennies, something even you didn't try to stubbornly argue was possible. That out-bound link will only confer one-fourth of the page's PR, not another third. And where will we get the one-fourth from? Again, it's not in the literature and there are no citations, but the answer is certainly implied clearly enough you should be able to figure it out.

What's that? Did you say we should steal a little bit of PR from each of the internal links on the page? (IMG:http://www.highrankings.com/forum/style_emoticons/default/smile.gif)

Cool. We'll take just a little bit of conferred PR from each of the one-third links, just enough to drop then down to one-fourth links, and lo and behold that will leave us just enough to come up with the necessary one-fourth for our added link. Instead of three links conferring one-third the page's PR, we get four links conferring one-fourth the PR. Our internal pages lose PR they would have otherwise received, but the math works pretty good and we're not in any danger of ending up trying to explain 102 pennies.

See, Michael, that wasn't so hard was it? It required a little bit of thinking, a little bit of digging beneath the surface to what is implied rather than baldly stated, but all in all, it shouldn't have been too painful. The three pages HAD to lose some of their PR for the fourth page to get its share.

Unfortunately, I doubt you're ready for the final step, and frankly, I just don't have the patience to walk you through it. Suffice it to say, as GBRD already implied, those three internal pages don't exist in a vacuum and the iterative algorithm if far from done with them. They link to other pages, probably mostly internal ones, and because they received less PR, only one-fourth instead of a bigger one-third, they now have less PR to pass to their targets. And their targets will then have less to pass to their targets, which eventually will probably lead all the way back to our page with three internal links and an added out-bound link that started this whole mess. The algorithm is said to be iterative, but in truth, it's almost incestuous. We walked very slowly through the process of stealing a little PR from three pages so we could confer a little PR to a fourth page, but over the course of several iterations of computing PageRank those four pages are going to affect countless more, almost inevitably leading back to the source page. All of the PR passed through the out-bound link, including the PR the target page will eventually pass to countless other pages, had to come from somewhere.

I know it boggles the mind, Michael, but a probability statistic can't ever exceed one hundred percent. Flipping an honest coin will result in fifty percent heads and fifty percent tails. Loading the heads side can't ever result in sixty percent heads and fifty percent tails because that exceeds the realm of possibilities. If you want to increase the chance of landing heads to sixty percent then you MUST be willing to decrease tails down to forty percent. It's simple math and, no, I'm afraid I don't have any citations or links to make it any simpler.

PageRank is nothing more than a coin with several billion sides.

You can't increase the chance of hitting any one side, which is exactly what PR was designed to predict, without simultaneously decreasing the chance of hitting other sides. You don't have to like it, you don't have to understand it, and you can stubbornly insist all day long that the original literature doesn't explicitly say it, but that's simply the way probability curves have to work.