How to Estimate How Many Links You Need
Doing some work for a client with a fairly new site recently, I noticed their home page has a toolbar PageRank of 7, and they achieved that almost entirely by having a sitewide link on a PageRank 9 website. Finally, now that have I seen enough cases like this, I think it’s possible to estimate the answer to the question “How many links do I need?” for many situations. This posting will detail the anecdotal data I’ve seen, and then give tables that can be used to estimate what a page’s toolbar PageRank will become if you obtain a link from a highPageRank site, and also how many links of which type you need to move up one level of PageRank.
If you’ve not read my previous article on Search Engine Land “What is a Link Worth?”, there is some good background in it, including a useful table of how many average links one needs to reach each level of PageRank.
One flaw in that analysis is, no one knows how much PageRank the “average” page on the web has; it’s definitely somewhere between toolbar “N/A” (which I consider to be absolute zero although it can also mean “Google is not telling”), toolbar “0″ (which I consider to be fractional anywhere between zero and one), and toolbar “1″ somewhere.
So the chart in that posting is only useful for rough purposes when planning out a general linking campaign; it can’t be used to do exact situational calculations. We’ll make an attempt at doing so here.
Anecdotal Situations I’ve Observed
1.) (As noted above) a new PR7 site whose links are almost exclusively from a sitewide link on a PR9 website.
2.) A PR5 website I know of obtained a home page link from a PR8 website and almost immediately moved up to became a PR6 website itself.
3.) A website with almost no links obtained a sitewide link from the PR6 website and became a PR4 website almost overnight.
Why Getting a PR8 Sitewide Link is Like Getting a Single PR9 Link
You’ve probably noticed that if you have a PageRank 6 website, the next level of pages down on the site are probably PageRank 5, and the next level down from that are PageRank 4, and so on. This is not always the case (for instance, one popular post on this blog titled “Google’s Secret Ranking Algorithm Exposed” is a PageRank 4 as of this writing, while the blog’s home page is a PageRank 3) but it’s generally the case. So when you obtain a sitewide link from a PR6 website, you’re essentially obtaining one PR6 link, maybe 10 or 11 PR5 links, and 50 PR4 links (perhaps).
If each level is worth roughly 1/5 of the previous, then a sitewide PR6 link is equivalent to:
1 PR6 link (the home page) + 10 2ndlevel PR5 links (=2 PR6) + 50 thirdlevel PR4 links (= 10PR5 = 2 PR6) = 5 PR6 Links = 1 PR7 link.
Looking at the anecdotal cases, it does seem to me that obtaining a sitewide link is akin to obtaining a home page link from a site that is one PageRank level higher.
The Value, in PageRank, of Obtaining a Single Link
Clearly, that depends on your existing PageRank, and how much PageRank you will get from the target of a linking request. Using the anecdotal evidence above, I created the following table and calibrated it against those individual situations; you can look up what your pages existing toolbar PageRank is, and then look over and see what the pages PageRank should be after obtaining a link from various other pages.
So, for instance, if you have a PageRank 3 page, and you get a PageRank 3, 4, or 5 link, don’t expect to see anything happen, but if you get a PageRank 6 link, you can expect much of the time to move up to PageRank 4 (see – row 3 represents a starting PageRank of 3 – the “PR 6 Link” column represents a PageRank 6 link, and their intersection is a 4 which means, you will end up with a PageRank of 4 after obtaining the link.)
Remember, this methodology is imperfect, and Google’s toolbar PageRank only reports integers – you could appear to be a PageRank 3 page, but really be a 3.99, and one PageRank 2 link might put you over the edge. Also, how much PageRank “flows” depends on how many links are on the source page.
But what you see in the table is typical based on my experience. You can see that for our anecdotal PR5 site that recieved a PR8 link, that is indeed enough to push it into PR6 status.
If what you’re looking at is a potential sitewide link, simply add one (i.e. the effect of a sitewide link, as we’ve shown above, is roughly the same as the effect of a single link from a site of the next higher PR level – so if it’s a sitewide link from a PR7 site, look in the PR8 column for what your final PageRank will be).
You can see from the table that these predictions aren’t too interesting unless you’re talking about highPR links; if you have any anecdotal evidence of your own either supporting or refuting this table, please comment below.
How Many Links You Need to Move Up One PageRank Level
Remember, in the article referenced above, the conclusion was that each toolbar PageRank level is roughly 5.14 times harder to reach.
Based on that assumption, and the table above, I generated Table 2, which does not show how many links it will take to go from zero links, but instead, how many links will it take you to *move up* one PageRank level. I often find that with most clients, their overall site needs to move up either one or two PageRank levels; this table can be used to estimate either case (to estimate two moves, just use it twice):
The left column represents a page’s starting PageRank; the table represents how many links of which type are required to move up at least one level. If a PR1 page obtains a PR10 link, obviously it will move up a lot more than one level, but you’ll only see a one in this table for that value, since that’s all it takes for the motion to occur.
So for instance, a PR3 link needs 5 PR4 links in order to become a PR4 link. Each level requires 5 times as many since their value is 5 times less.
An astute observer will note that there is one exception – there is a diagonal of “3″‘s in the table. These were originally “1″‘s based on my anecdotal evidence, but represent the borderline case where you may be a 5.01 or a 5.99 – by having “3″‘s in the table in those spots, I’m splitting the difference and it should be reasonably correct most of the time. Yes, it’s cheating a bit but the borderline cases really required it.
Checking the Table
If you take the number of links at each level and accumulate them for the “PR0″ and “PR1″ columns, you can see how many “PR0″ or “PR1″ links it will take *in total* to reach each level from nothing (i.e. for PR3 for instance, you just add up how many it took to reach each previous level). Table 3 shows this:
Since the *average* page on the web ought to have a PageRank somewhere between “PR 0″ and “PR 1″ (warning: major assumption!) then we would expect that, on average, if you were obtaining *average* links, the number of links required would be somewhere between these two columns.
If you compare these two columns to the table in my previous article on Search Engine Land “What is a Link Worth?”, you’ll notice that they fall roughly equally on either side of it, just as we’d expect! This is a pretty good check and is enough to convince me that the two tables above have good predictive value.
Full disclosure: if the average page on the web instead has a PageRank somewhere between “PR N/A” and “PR 0″, then this argument completely breaks down and these columns are probably off by at least a factor of 5. Ultimately, I think the only way we can gauge the accuracy of the tables above is by people actually using them and reporting their results.
Conclusion
Of course, links have more value than simply PageRank – anchor text is highly valuable in and of itself, and if there’s any reality to the concepts of “trust” and “authority”, then there are many factors that need to be taken into account when planning linking campaigns.
From a pure PageRank perspective, these tables are very rough tools, but they have at least been calibrated against some realworld data. Go ahead and try them out and give some feedback below; perhaps they can be improved, or maybe certain aspects are *way* off – please share any findings.
As always, your mileage will vary – you might plan out a linking campaign and assume that X number of a certain type of link will get you to PageRank Y, and then it turns out in the real world it doesn’t, and your boss or client will then be disappointed, so *use these tables at your own risk*. You can always pad your estimates like Scotty here to be safe
But in the land of the blind the oneeyed man is king – these tables are better than nothing presumably! Any feedback below would be much appreciated.
Great information Ted. I’m glad you found me on Twitter. have been learning a lot from your articles. Keep up the good work. I am in the process of optimizing my companies site and I’ll be sure to report back when my page rank and SERP are better.
Great article and some good food for thought. I wonder though if you looked at:
the effect of the number of outbound links on each page and how this would alter the link juice received form that page. If you had a link on a PR5 page with no other outbound links vs a page with 75 outbound links. Surely this would dilute the strength of the link and affect the chart? Did you assume an average a “x” numbe rof OBL’s?
Wouldn’t you assume that after “x” amount of links from one site, they start to lose the affect. If you had 1,000 site wide links from one domain that is not enough diversity. Can we assume that there is a certain saturation after “X” many links from one domain that will help you?
I will reread this post and the related other posts and check back.
Thanks for the great insights.

Kevin – great questions and points.
My thinking was that these would be links from pages that have “average” number of links on them (i.e. they would clearly vary, but their average should resemble the “average” page on the internet).
There are two studies that have answered this question – one in 2006 by some folks at Cornell titled
The Portrait of a Common HTML Web Page:
http://webseer.sourceforge.net/papers/doceng06.pdf
In it they found that the average page had:
41 links
10 of those were outside of the domain
2.8 of those were to other subdomains
29 were to the same domain or document.
More recently, the SEOMoz folks did an update of LinkScape’s crawl:
http://www.seomoz.org/blog/november2011linkscapeupdate
…and found that the average page had:
77 links
65 of these were internal (presumably same domain or subdomain)
12 were external (presumably outside the domain).
This matches pretty well except it shows that the number of internal links on pages has increased in recent years.
Yes, any PageRank you get would be divided out among the various links on the page, so a linkladen page would get you less PageRank. But on average if you got 1,000 links, that should come out in the wash, unless you got them all with the same approach – this is a good argument for having a multipronged linking campaign.
As far as the value of additional links from the same domain or subdomain being less and less, I’m pretty sure there is a Google Patent application that mentions that concept, also SEOMoz’s correlation data showing that rankings correlate higher with links from unique domains than total links pretty much confirms that. As to what sort of decay there is, I have seen zero data. It could be that links #2 through #1000 from the same domain are worth the same, or maybe they’re worth progressively less. Good question.
Hi Ted,
I can see you have put a huge amount of effort and research into gathering this data. It makes for fascinating reading so thanks very much for that.
All that I would add is that Google only updates the published PR for a page every now and again (12 times a year I believe) so monitoring SERP ranking changes may be a more immediate indicator of change than checking published page rank.
I can see your charts being of help in the analysis of the link profile of say a #1 ranking competitor to help clarify what needs to be done to overtake them.
I’ll be giving them a try out. Thanks.
Pete
Just wanted you to know that my boss and I read your blog regularly, and have been doing so for quite some time. You have very intriguing views on SEO, and I definitely can appreciate the ideas you write about. Thanks for continually sharing your knowledge.
I think your info is correct, or simply very accurate. I agree with the facts you are mentioning here in this article. I have seen similar effects with some of my web properties. Thanks for sharing this info with seo community.
Thanks for that. I always wondered how many links you need for pageranking. But I suppose for general seo the more the better?
More than interesting. I have to study a little bit to comprehend everything was said here,
Thank you! That’s a lot of very useful data.
Note to all – I originally published this with tables reflecting a factor of 5 (i.e. 5.0) increase in PageRank at each level.
I have now updated tables 1 and 2 now to instead take into account a factor of 5.14.
It makes small differences at low PageRank levels but increasingly large ones high levels, so the tables should be more accurate now.
Hey Ted great break down. I am printing out the charts right now. So now I need to find a PR 9 link! ha
I assume you’re talking about dofollow links?
Hmm. No, just links, overall. It’s really unclear whether and when Google respects nofollow and when it doesn’t, that is an unsolved mystery as far as I can tell – there have been many mixed messages over the last few years from Google on this.
Great post!
I have to agree with the charts and what you say.
Have you found that links from similar categories / topics have a higher boost to rankings ie. a link to a poker site from another poker site has a higher value verse a poker site with a backlink from an unrelated topic? That’s assuming both sites providing the back links had the same page rank.
Also I’m intrigued by the comments from you and Lewis about the nofollow links. I just had assumed you meant all dofollow in the post but good to see your feelings on nofollow.
Someone necessarily help to make significantly posts I would state. That is the first time I frequented your web page and up to now? I surprised with the research you made to make this actual publish incredible. Wonderful process!
This is interesting. I’ll will test this on my 2 sites and let me see the result when next pr updates.
I think that this scheme may work for getting a PR3 (maybe a PR4 website), but from there on things get really difficult and in order to be ranked you need a honest website with premium content that’s cited by other established website. As expected, content is still the king!
I wonder if this is relevant today!
very interesting post, thank you