Jonathan Leger – SEO And Internet Marketing Blog Internet Marketing Blog

5Jun/12Off

Link IP Diversity – How Important Is It?

Geographic diversity. It's a requirement for getting links to your site, right?

That is, if you want to rank in Google, then you need to make sure that you get links from all over the place, because Google won't rank a site if all of its links come from web sites that are too "close" to each other. Right?

We've all read that. I've believed it for years. I started believing it when I tested the theory, and the results seemed pretty conclusive back then.

The SEO "gold standard" for determining whether or not two web sites are geographically diverse has long been this: does the web site you're getting the link from reside on a different Class-C IP address from the other sites that you already have links from?

 
What's A Class-C IP?

If you're already familiar with how IP addresses work, and how to determine if two IP addresses have different Class-C addresses, you can skip this section. If not, then read on.

Every web site has a unique "address" its assigned in addition to its domain name. It's a set of 4 numeric values connected by periods. (For instance, the IP address of this blog is 72.34.63.94) Each number in that address is a "Class". The first number (72) is the "A" Class, 34 is the "B" Class, 63 is the "C" and 94 is the "D".

So if you were to get a link from my blog to your web site, and you also got a link from any other site hosted on the same IP address (even though the domain name is different), SEO common knowledge says that Google won't count the additional link from the same IP address very much. At the very least, you want the link to come from a site whose "C" Class IP is different.

That is, you don't want to get a link from a site on 72.34.63.94 and also one whose IP address is 72.34.63.50 (those have different "D" Class numbers, but the "C" class is the same). But a link from a site on 72.34.63.94 and another link from a site whose IP address is 72.34.10.12 is good because it's "C" Class address number is different from the address of this blog.

 
The Theory: Why IP Diversity Matters

Of course, it's inevitable that you'll get some links from sites on the same IPs or IPs whose "D" Class is different but whose "C" Class is the same. The idea, though, is to get as many geographically diverse links as you can, otherwise Google will figure you're up to no good and not rank your page.

After all, if the sites you're getting links from all have the same or similar IPs, then they're all sitting on the same server (or on servers that are very "close" to each other, since geography plays a big part in what IP address a server or web site is assigned). That's not a "natural" linking pattern, and so it would make sense to discount those links.

Or at least, that's been the theory.

 
Setting Up The Test

With all of Google's major algorithm changes lately, I thought it was time to retest the IP diversity theory and see if it still holds true. So I took the results from my anchor text optimization case study and did some additional backlink analysis on them.

To keep things focused, I decided only to analyze the sites whose index page was ranking for the keywords (that is, if mydomain.com was ranking, I kept it, but if mydomain.com/page.html was ranking, I left it out). I also decided to include in the analysis only those sites that had between 100 and 5,000 different domains linking to them. I did that for two reasons:

1. My customers tend to optimize their sites for low to mid-competition keywords that you can rank for with that number of links.

and

2. The more unique domains a site has linking to it, the higher the odds are that the IP addresses will have to be duplicated in the links. Consider the English Wikipedia site, for instance (en.wikipedia.org). It has some 1.7 million different domains linking to it. Those 1.7 million sites are hosted on about 557,000 different IP addresses, giving it an IP diversity of about 32%. But in those 557,000 IPs there are only 142,000 or so different Class-C IPs. That's only an 8% diversity. For a site with fewer links, 8% is way below what most people who do SEO would recommend. However, for a site with that many domains linking to it a low overall diversity percentage is inevitable. After all, there are only so many IP addresses allocated to English speaking countries that are likely to be linking to the English Wikipedia.

So that was my rationale, and about 3,500 ranking pages from 35 highly diverse topics met the criteria I used for this test.

 
What The Data Says

When I ran the data I fully expected to see some high percentages of unique Class-C IPs linking to the top ranking pages, and overall that's what the data shows. Here's a table of the overall results for the 35 keyword topics. The table shows how many sites meeting the criteria were ranking in the top 10 for their keywords within the topic, what the topic is, what percentage of the ranking site's links to its home page came from unique Class-C IPs, and what percentage of the ranking sites links to any of its pages came from unique Class-C IPs:

sites topic Unique IP-Cs (Homepage) Unique IP-Cs (All pages)
381 web hosting 53% 42%
296 ecommerce 59% 48%
258 shopping cart solutions 60% 56%
89 home improvements 60% 50%
221 email hosting 63% 52%
28 b2b ecommerce 63% 47%
283 web design 64% 59%
109 plumbing 65% 60%
176 pharmaceutical 67% 58%
204 virtual server hosting 68% 54%
95 government and trade 68% 46%
110 human resources 68% 57%
250 online stores 68% 60%
253 transportation 68% 62%
114 b2b 69% 57%
158 office supplies 70% 66%
203 agriculture 70% 61%
68 business travel 71% 62%
242 business 71% 53%
209 technology 72% 57%
89 biotechnology 73% 60%
118 business management 73% 57%
18 industrial goods 74% 40%
162 home based business 74% 61%
211 work from home 75% 63%
203 management 75% 59%
108 semiconductors 75% 60%
128 chemicals 75% 61%
297 software 75% 56%
103 aerospace 75% 59%
55 office management 76% 66%
131 internet security 76% 55%
34 food and beverage 77% 64%
19 office backup 79% 72%
125 startup 83% 61%

You can see that the lowest percentage of links coming from sites on unique IPs was 53% to the home page and 40% to the entire site. That's pretty high (certainly higher than the English Wikipedia's 8%). That means there's lots of IP diversity, at least overall.

What surprised me, though, is what I saw as I started to dig into the individual ranking sites. There were 128 examples of sites ranking with less than 40% IP-C diversity -- some with far, far less.

For example, pharmaceuticalrepjobs.com is ranking in the top 10 in Google for "pharmaceutical representative", "pharmaceutical rep" and "pharmaceutical representative jobs." According to KeywordCanine.com (which I used to gather the data for this test), that site's home page has 375 different domains linking to it -- but all 375 are only on 16 different IP addresses (13 different Class-C IPs)! That's barely 3.5% diversity. The entire site only has 377 different domains linking to it from 16 different Class-C IPs as well, so it's not the strength of the entire site's linking pattern that's the cause for the ranking.

Meanwhile, the great majority of the other sites ranking for those keywords have much higher numbers of unique domain backlinks with much greater IP diversity.

That site wasn't alone, either. There were dozens of examples of sites ranking for their keywords with ultra-low IP diversity. Some had a few dozen backlinks, some had hundreds, others had thousands of unique domains linking in. Some of the sites were exact matches for the keywords (which I also explored in my previous case study), but others weren't. I really couldn't find any specific correlation that fit for all (or even most) of the sites who seemed to be "getting away with it."

 
The Take-Away

So what can you take away from this experiment? My conclusion is that, although there are some exceptions to the rule, they only make up a small percentage of the ranking sites (less than 4%). Most of the ranking sites have a much larger amount of Class-C IP diversity in their backlink profiles.

In fact, out of the 520 sets of keywords that were included in the data set, only 14 keyword sets had less than 40% average Class-C IP diversity in their backlinks, only 39 had less than 50% Class-C IP diversity, and only 124 had less than 60% (that's just over 23% of the results). So for three out of four keywords, the average Class-C IP diversity for the results was at least 60%.

That seems to indicate that the geographic link diversity is still important, even if there are some notable exceptions to the rule.

If you have any questions or insights, please leave a comment below. Your feedback is always welcome!

Oh, and one last note: if you found this post beneficial, please share it using one of the buttons below: