NOTE: After publishing the initial version of this post I realised I had a concentration lapse and you actually need the paid API to get the full automation from this tool. I have updated the post now with a work around for those without the API that requires an extra step to export the data from OSE. Sorry about that, folks!
A couple of months ago I got pulled in on a project for a new client, who admitted that their previous 'SEO guy' had bought them some links. Not a lot of links (they didn't have a lot!), but it seemed that had been the extent of their SEO efforts. One of the things we needed to do was evaluate what amount of damage might have been done.
What I'm going to share today is a tool I've built that automates the process I used to evaluate how our clients link profile looked and whether it stood out against other domains in its niche. I'm going to show you the real data I saw for that client (who we knew had bought links) and how you can use the same tool to identify domains that have maybe bought links or other anomalous link profiles. Or you may just want to use it compare your domains link profile to your competitors, and see if anything stands out.
1. The Process
I'd previously encountered a nice way to break down a link profile via my Distilled deskmate, Dave (@dsottimano), and subsequently discovered that Dr. Pete (@dr_pete) had also introduced such a technique in a previous SEOmoz post. The premise of the method is that you break the links to a domain down by the Domain Authority of the linking root domains, and measure how many links are at each level. Here is how Distilled.net's Link Profile looks:
You can also do it via Page Authority, but in this post I'm going to concentrate on DA.
As Dr. Pete discusses in his post, these graphs can in and of themselves be very informative. However, for my purposes I needed to go further: I needed to compare the link profile of our new client to other companies in the same niche and see how far off our client was. As a side note, it is important that you compare like for like when looking at link profiles - different niches tend to have different link profiles.
So, it seemed pretty simple: we'll just chuck the link data for several competitors and our client all on the same graph and all our problems will surely be sorted, right? Here is how the data looked:
Hrm, ok. This wasn't particularly helpful!
The problem is that different domains of different ages and popularity have different numbers of links. You cannot really see much from this graph, other than how many links each domain has compared to each other, which has far simpler ways to visualise! However, if you normalise this data, accounting for the total number of links by instead examining the percentage of linking root domains each of these domains has for each DA level and we get something quite different:
BOOM! Can you guess which one is the client? The blue line stands out like a sore thumb, it doesn't fit to the natural curve of the others. Those of you with sharp eyes will notice that we had less overall links and so with far fewer data points you'd expect the graph to be a bit more jittery than the others. However, the red line shows a competitor with a not dissimilar number of linking root domains and they seem to fit pretty nicely to the curve.
You can see there is a spike where all of a sudden a whole bunch more root domains in the DA 30 bin (relating to a DA of 25-34) are linking in; domains in that range are common candidates for the positioning of paid or spammy links.
If we can spot this so easily, then it seems plausible that Google can too.
Not convinced that the client's link profile is that far off from the others? Let's go wild and do some maths (sorry, I'm British, we do maths not math)…
I did the pairwise Pearson's correlation of the domains, meaning I compared every combination of them and noted their Pearson co-efficient. Using the built in PEARSON() function in Google Spreadsheets you get a number between 0 - 1 to indicate how closely correlated two sets of data are, with 1 being a perfect correlation. I averaged each domains correlation coefficients against the others to produce a graph of how each domain's link profile correlates with all the others on average:
Oh oh. Not looking an better!! This graph isn't perfectly scientific, but it is often both practical and illuminative.
Using the link profile graph, you can begin to identify any anomalies which you think may warrant further investigation, which I go into in section 3 below. You won't always get such a clear cut situation as this, but fairly often you'll turn something up.
Firstly, let's make this all a little easier with the power of the Linkscape API and Google Docs...
2. The Tool
The tool I've put together is built in Google Docs, so it is easy and free for you to make a copy for yourself. The link to the Google Doc is here:
You'll need to be logged in to a Google account to access it; once in go to the File menu and select "Make a copy" to create your own private copy of the tool you can use.
Unfortunately, only after finishing the tool did I realise you can only full automate the process with the Paid API, so with the free API you have an extra step. Please follow the relevant instructions.
Free Account
With the free API you can only get 3 linking root domains via the API, but you can get all you need from OpenSiteExplorer.org. If you aren't registered, get a free community account. Go to OSE and enter your domain and let OSE do it's thing. Then switch to the "Linking Domains" tab at the top:
Spot the "Download CSV" link on the right? Click that and wait a moment for the report to process so you can download it.
You'll find 3 sheets; one named 'Data' and one named 'OSE Importer ', you can ignore the 'Config' sheet. Go to the Data sheet where you'll find spaces to enter a list of domains; because you've not entered API details in the Config sheet nothing will happen - it is waiting for import data from OSE.
Open the CSV file you downloaded from OSE in either GDocs, Excel or another spreadsheet app. You'll probably find the list of DA figures starts in cell B2; copy that whole column without the header (B2 down to the last entry). Now open the 'OSE Importer' sheet and paste the data under the relevant domain name.
Repeat this process for each domain name you want to analyse and then return to the 'Data' sheet to see the pretty graphs and analyses.
Paid API
You'll find 3 sheets; one named 'Data' and one named 'Config', you can ignore the 'OSE Importer' sheet. In the Config sheet you'll find a space to enter your SEOmoz API details; I use the Links API and only the free elements so you don't need a Pro account. However, without a Pro account you'll be limited to 1000 root domains, so if you want to run this analysis on bigger domains bear that in mind. If you don't know your API details then you can grab them from the API page. You can also select how many root domains you'd like to include (todo small: if you are looking to investigate very large domains (above 10,000 linking root domains) then Google Docs will struggle to manage without using the APIs pagination facilities which I've not included here).
Now go to the Data sheet where you'll find spaces to enter a list of domains; do so without including the https:// portion of the URL. The graphs and data should begin to automatically populate the spreadsheet, and now if you scroll over a bit you'll find your link profile graph waiting for you.
3. Uses
Awesome, so now you can profile various domains links, what can you actually do with it?
Firstly, a few points to bear in mind:
- Not all anomalies are paid links or something else dodgy
- It can be fairly common for multiple domains in a niche to be employing dodgy tactics. Be aware.
- Basically, the tool will point you in the right direction, but you need to investigate.
So, I think there are probably a lot of uses for this sort of graph, so please make suggestions in the comments. I'll get the ball rolling:
- Identifying possible link buying or other spam linking.
- Verify the link health of new customer.
- Discover powerful link strategies your competitors are using (i.e widgetbait, link-potent PR etc).
- Just for the sheer fun of seeing how you domain fits in against the others in your niche.
Things to look for on the graphs:
- Is there a disproportionate spike in the number of low to mid strength inbound links?
- Is a particular domain more successful in attaining higher quality links on average?
- Do any of the domains not fit to an approximate bell curve type profile?
4. A Quick Test
As is known, GoCompare.com previously bought links, and not good quality links like some of their competitors were maybe doing, but really low-quality spammy links. So, let's see if we can pick up those remnants (the problem with mass buying of links is it is hard to clear them up) with the tool. I threw GoCompare.com into the tool along with competitors of theirs that I grabbed from the first page of results for the search 'car insurance': morethan.com, admiral.com, elephant.co.uk and lv.com.
Here's the link profile:
We can see that GoCompare.com (blue) and LV.com (purple) are both slightly outlying. Also note that the GoCompare spike occurs around the DA 20 range - indicating low-quality links as was reported. Lets take a look at the average correlations:
We can see that whilst GoCompare correlates well with the others, it is the furthest off. Whilst not conclusive, it seems to suggest that they are working on it but still have a bit more work to do. With this example, we again knew what we were looking for, but hopefully you can see that if we were just looking to confirm suspicions or do a bit of research, this tool can help you along the way.
5. Wrap Up
The tool I've presented is not supposed to be an in depth analysis, but as part of your armoury it can allow you to very quickly and very easily evaluate a domains link profile and hopefully gain some insight into where to dig deeper.
It seems that SEOmoz have got yet more awesomeness up their sleeves in the coming months, including rolling out an improved Domain Authority algorithm that is going to more closely correlate with the SERPs. That change will likely improve the potency of this tool yet further.
If you have suggestions of other applications or similar techniques you are using, please do share in the comments! :)
Hi Tom, liking the idea behind this and it could be particularly valuable in a pitch process or to start looking for past penalties when deciding what to do about a clients existing paid links from a previous agency: use the sheets, do the research, look for 301 redirects (or other signs of past penalties), make an assessment, rinse and repeat.
With that said, I think it also highlights the difficulty in Google trying to use activity like this to nail down absolute cases of "paid" or "dodgy" link profiles and why jumping to conclusions without any sort of manual review would be a very bad idea. I think your comment:
"Not all anomalies are paid links or something else dodgy"
is an extremely important one and maybe one that doesn't come across as clearly as possible with the title. I think the other important thing to bear in mind is the anomaly situation. The "everyone else is doing it" excuse isn't always the right approach. In some of the most competitive niches I'm starting to see the anomalous profile belonging to the "cleaner" link profiles and it may indeed be a sign of someone doing it right rather than breaking the rules... food for thought there.
In any event, many thanks for sharing the tool, it automates a good first step in the process and helps figure where to start digging so I'm all for it! But I think the fact that this is only the first step in the process might suggest quite strongly why Google still requires manual review, spam reports, etc. It takes a trained eye to decide whether anomalous means "doing it 'right'" and "being spammy".
Hey Sam,
I couldn't agree more. This should only be a starting point to guide more in depth research (digging into those links a bit etc.). I think you are right - there are a tonne of reasons why certain link profiles may not be the same as the others in their niche, and often it may be because they are doing something awesome, or are the only person not doing something dodgy.
There never will be a silver bullet for turning up sites that are doing something spammy; everything on both sides is constantly changing and many black/grey techniques have a corresponding white hat technique that will create a similar looking profile.
Thanks for the feedback. :)
I often use the anchor text distribution analysis feature of Open site explorer to determine paid links. I Compare the anchor text distribution (setting:show anchor text 'phrase' for links to 'all pages on this root domain') of my site with my top competitors. Often first page of the report tell me all the story. Generally all websites will natural link profile have maximum no. of linking root domains containing their brand/website name or its variations as the anchor text. That's how people naturally link out. Now if maximum no. of linking root domains target your top industry keywords (esp. with exact match keyword phrase), then their is something to investigate. Click on the little '+' button next to the anchor text term to determine the domains that are linking out with that anchor text. Now look for following charactersitics while evaluating suspicious websites:
1. Try to determine the website purpose. Link farms generally don't have any purpose. So they write about every topic under the sun and link out to every tom, dick and harry.
2. Check out where the link is placed? How the anchor text is used? Does the use of anchor text looks natural?
3. Is the website really made for visitors? Do you see any signs of user engagement (like commenting, tweets, facebook likes etc).
4. Can you justify the link placed on the website? For e.g. a website on php linking out to cats food website.
5. Is contact information available on the site?
All this make look like a tedious task but it is not. If you keep these points in mind you can detect paid & SEOed links pretty quickly.
Himanshu your post deserves its own article really fantastic comment mate. Your points are very explained and up to mark. Thanks for sharing some instant tips to track suspicious website.
Nice work guys.
I made a suggestion that this feature should be built into the SERP analysis report too if possible :-) -
https://www.seomoz.org/blog/seomoz-feature-preview-the-serp-analysis-report#jtc138419
Tom - I'm getting an error too, is the tool still functional?
Is it bad that I plan to use this tool to test better link-network usage? Muauhaha. But on a more serious note.. nice work guys this looks awesome.
You thinking about testing out potensial domains to acquire links from? Thats a bloody good idea if you ask me :) Why didnt i think of that?...
Sorry everyone, I had a bit of a brain-fart and missed that whilst you can get 1000 links from the free API you can only get 3 linking root domains. So you need the API for the full automation.
However you can export the domain information as a CSV from OSE and paste the values into the tool which will then do the graphs and correlations.
I'm just working the adjustment into the post. Sorry!
I have a pro account but I only get 3 root domains in my sheet.. What am I doing wrong Tom?
hmm same for me.
Hi Björn,
There is a difference between Pro accounts and the Paid API (https://apiwiki.seomoz.org/SEOmoz-API-Pricing). The Free API is very powerful, but unfortunately it does limit the number of linking root domains (which I forgot - I thought it was any 1000 links).
However, I've just updated the post with a work around to export the data from OSE. It is an extra step, but you can still run the report in a couple of minutes, which I think is still really quick for what it tells you.
I am having the same problem. Only 3 root domains, with free API account.
Apart from that, thank you for this highly interesting article!
Hi Tom,
Thanks for the update, I have now tried the csv approach ;)
The results are very good. Thank you for the update in your article.
Interesting, all my competitors seems to have very similar link profiles.
But no big surprise, of course the distribtuion is peaked somewhere at medium authority values. Still one might be able to deduce slight differences and their potential cause as outlined in your article.
Seomoz should include your tool in their suite.
Dr. Tassilo,
Really glad you've found it useful! :)
Yes - most link profiles have an approximate bell curve shape, which makes a lot of sense. But by comparing against others you can hopefully find anything that stands out.
Thanks for the feedback. :)
Also, I think we should change the drop down so it's links to the whole domain? It will default to just links to that page. - jon
Hi SEO Consult,
Unfortunately, that doesn't change the download. The CSV approach will only look at links to the homepage for the moment.
Damn they've thought of everything! Ok thanks Tom - Jon
Actually I've just checked and I am getting mor results in the CSV export when choose "Links to this Domain" rather than "Links to this page".
I tried it out and it works like charm! Fantastic. Thanks Tom :D
If my link correlation looks like this, is that Good or bad?
my domain .73
next .67
next .58
next .55
next .57
It shows only 3 root domains instead of 1000 , as I have a free account .
It's only showing 3 domain links no matter what URL I put in (and i've tried bbc.co.uk). I presume I've set everything up correctly or it wouldn't show anything but it's a little odd!
I'm also getting a maximum of 3. Anyone have any ideas?
Hi,
I've updated the post for this problem. Please see https://www.seomoz.org/blog/link-profile-tool-to-discover-linking-activity#jtc154084 for details!
Great stuff, Tom. Really happy someone expanded on this concept (and I didn't have to do the work ;) ). When you think of the Google spam algorithm as essentially a giant pattern-matching machine, it really does make sense to look at how and why sites stick out. There are good and bad ways to stick out, of course, but this analysis is a great start to digging deeper into the problem.
Thanks, Dr. Pete! Yeah - think this tool can be really illuminative, but you have to dig a bit deeper to confirm.
Ok, I wasn't hopeful, but I went through the steps and it completely 100% confirmed our thoughts that our backlink profile was abnormally spammy. We already knew that I guess, and are actually trying to eliminate paid links that have caused penalties, but it really helps to know that our backlink profile sticks out like a sore thumb compared to our competitors.
I think the people that thumbs downed you just couldnt get it to work.
Thanks
clinthend,
Thanks for this great feedback! I'm really glad that you found the tool useful. Hopefully, you can use what you've learnt to start to sort that link profile out a bit! :)
Tom,
I just set this up and I love it. This really helps paint an overall picture of your backlink profile.
Thanks for posting
Tom, thank you for a really interesting blog post, appreciate the insight and free gear. I especially loved this sentence, "If we can spot this so easily, then it seems plausible that Google can too.".
Amazing post as usual!!! Thanks for sharing the tool too - should be able to finally get my head around using scope and filter within the links API!
How do you know the 'other' sites are not buying links
within the 'linking profile'?
I think it's dangerous (albeit safe in the beginning) to follow
the crowd. google wants to determine what people like to see
and how sites link 'ntaurally' to each other. but ehy don't have
a'pure' web lab to do so. so their data is already skewed.
Anywya, great material nonetheless.
Yves Baggi
Love it!
Looks good Tom, I have downloaded the spead sheet and will test it out, I think what also works well with these link profile's is where you can do analysis on the keywords anchor text values used via specific sites over a time range, it you see a huge influx in keyword rich anchors for specific terms on high power domains it is an indication something is going on which needs to be investigated.
You also might want to check out this post (mine) on SEOMoz a month or so back...
https://www.seomoz.org/blog/the-wikipedia-model
It provides 3 different tools for detecting link anomalies using Wikipedia as the control.
I'd love to see a combination of your approach and the stuff TomA has here - an overall "link profile analysis" tool would be super awesome :)
Russ - Thanks for sharing that. I had meant to link to your (totally awesome!) article in my post. :)
wow, cool tool. Thanks
This is a really awesome analysis. I will definitely be using your insights and your spreadsheet. Thanks!
Great piece of infromation. I'll try the software!
I woke up this morning hoping for a new tool to play with. Just spent 2 hours 'playing' with this, very cool.
Awesome tool. Just added 4 competitors and compared our link profiles to see how things were looking. Turns out that one of the biggest names / brands in our niche that we know have been buying links correlated at about a .65 while we sat at a 0.95. I'm thinking of sending a screen-grab and a load of other evidence to the webspam team but just can't quite bring myself to do it ;)
A .30 disparity is pretty large - but be sure to check out where those disparities actually are. This tool will show anomolies, but the disparities don't always mean someone is buying links.
In this day in age, you never know! Best of luck!
Mike
nice sheat, but it doesn't work for me. I entered the domain names (host names), the access key, the secret and 1000 root domains because I'm using a free account.
I removed also all whitespace. Below each domain I get after some short "thinking" the #Error message
Is there something I need to change?
It helps if you type them in manually rather than cut and paste. I had a similar problem and sorted it byt actually just typing the URL. not sure why but it worked.
I ran this a few times today with a couple of my clients.
Very important - you said, "basically, the tool will point you in the right direction, but you need to investigate." I found that exact match domains, especially those that are the home to several companies, are false positives. It makes sense. But it did lead me to some deeper digging on some competitors that weren't on my radar for paid links.
And that's what this worksheet was supposed to do. Big thanks!
Thank you very much for this. I just went to OSE and when I click on Linking domain it shows me also that among all the websites linking to my site, my own website is linking to my website also. Is this normal?
And why when I click on inbound links and want to filter only no followed it says no data available for this url? The website is https://villasdiani.com what is wrong:-(
I need to sit down and study this post as it is so helpful but also with such a good level of detail. Love the Google Doc work if anything I take away from this it is the nice graphs that Google Spreadsheets create that look well on websites. I am a visual fan and the nice bar charts and line graphs here really enhance the message. Of course the message is the charts I know but I think Excel would not have looked as well in the post.
Now I need to add this to my "SEO Toolkit / Bookmarks list"
Thanks...
Thanks Tom, but for some reason it's only reporting a very low amount of links for each domain.. Anyone else having the same issue?
It's probably Linkscape not having enough data. If you're working on smaller sites than you'll often run into this kind of problem. Always try more than one source. If Linkscape doesn't have enough data, maybe MajesticSEO has.
I'm having this problem too Netsociety. I'm getting 1 or 2 links per domain even for well established sites. The links all show as DA 80 or above, nothing lower. I regernated my API key just to be sure but same issue...
I'm having the same problem. I get 3 links for all the domains I put in, doesn't matter what the URL is or how big the site is. :(
I was looking forward to using the tool to help develop my link strategy (what I need to do, what others have done, etc.)
Same deal, exactly 3 links.
Same here. I get exactly three links each for all five domains. These are all fairly high authority domains with lots of data in OSE. Strange...
Ah, found the problem, thanks Tom: https://www.seomoz.org/blog/link-profile-tool-to-discover-linking-activity#jtc154084
Does not show all domains
Hey, great post! But i'm getting an error:
Request failed for https://lsapi.seomoz.com/linkscape/links/www.mydomain.com?AccessID=MYID&Expires=[cut...] returned code 401. Server response: invalid signature.
I posted the AccessID and the Secret Key from my PRO account into the config sheet field B1 and B2.
Im pretty sure everything is fine... may anyone can give me a hint?
Hi,
This is really very nice and informative blog and i love to read this informative post and i also wnt to some more knowledge about that is you can share then i am thankful to you.
I wanted to thank you for Link Profile Tool to Discover Paid Links ! Definitely enjoying every bit of it I’ve bookmarked to check out new things to post many thanks for this great post! I’ve been looking for this for a long time.
Great R&D - thanks Tom!
I initially wrote a giant comment, but then it turned into a whole article on my blog on whether Google can detect paid links. :-)
The short version is that your graphs made it clear that Google could statistically profile backlinks to detect artifical links. Given the data set available to them, and the quality of engineers, I woudn't bet against them doing a great job over time.
My big takeaway from this was to be very careful with how I do link building - thank you!
Dont work for me... i get an error
error: Request failed for https://lsapi.seomoz.com/linkscape/links/***REMOVED***<?> returned code 401. Server response: invalid signature. String To Sign: ***REMOVED*** You provided:'***REMOVED***=' (line 43)
This is an awesome post. took a 'copy of the tool' and played with it as well. Gave me some forward thinking for another tool I will try create (my excel skills are not that good though)
cheers
@seowestcp
Tom! Thank you for this tools. I think it's great! I've just tested in on two projects and it show great, valuable data. I'm insipired by this to develop this model with some other functions. Thanks again!
Thanks for the Blog, awesome..
I have a client who's main competition is absolutly smashing them on the the SERPs for all phrases, just because they are buying links through a large agency in Brighton who are sub-contracting the link buying to an agency in India. Nearly 50% of their links are paid blog entries and so black-hat.
My client would like to report them, but does not want to use their webmaster tool account, is their an alternative root to outing the offending domain?
Has anyone made an update to this one?
This is a handy solution, good for presentations. Why not have it as a basic MOZ tool? Thanks.
Does anyone else have a problem with the formulas working? I use the OSE importer page and everything is right on that page but it doesn't transfer to the DATA page gives errors?
Can someone send me a copy of a working sheet?
Thanks!
If my link correlation looks like this is that Good or bad?
my domain .73
next .67
next .58
next .55
next .57
Same story here.
Keep getting ERROR while trying to import data through the OSE API
what are other tools for link analysis and back link analysis ?
Nice Tool, I use it my every single day.
big thnks
Yeah - I agree with some of the comments above. I don't think it is really, shall we say "identifying paid links" as alot of it is just pretty graphs however it does give you an indication.
Thanks for sharing!
Awesome stuff :D
I have the PRO API but I am still only getting 3 backlinks in the DA columns all around the 80+ range.
I have even tried using the gocompare control and it still only shows 3 backlinks in the 80-100 range.
What am I doing wrong?
this will definitely help my links as they grow, what is the criteria to assertain a bad link ?
this is truely informative & intellectual stuff.. I was using some distilled & other ones docs tricks also.. Now I am wondering after privacy policy of google.. is it that safe to play with google docs for SEO & Social Media Purpose? May be I am too pessimist on this issue nit sure :(
Tried this tool but I have errors coming from time to time.
pretty cool tool to play around with!! thanks!
Thanks - it's a great tool. Have run my site vs some competitors with some very interesting results - most correlate very well but it's super easy to see those that don't..
thanks tom, brought such awesmoe useful stuff!
I have added the seomoz api info. But when i added the domain it is giving me error in spreadsheet. Let me know If anybody has faced this issue?
I'm having the exact same problem... am using a PRO account, and have enteredt the member-xxx id and key and the error message that I get after typing in the url gives 401 authentication failed...
Good ideology, I must say this is very unique article I have found yet! Great work.
I sometimes think you guys are smarter than Google. I love graphhs and this is a nice one. Thing is that I used it for a website I work for and 4 competitor sites. It turns out that the two sites with the lowest av. correlation have the first positions in SERP. Both have an almost opposite profile. The site in position 1 has also the highest % of high DA links. The site in position 2 has low % high DA links but much higher average low DA links. Very akward.
At first sight this does not show that Google uses these patterns as shown. There must be more factors.
- Are links coming from sites related to my business sector
- History of linkbuilding
- What stategies do I and my competition use for linkbuilding
For a quick insight in link pattern this is a very useful tool but at the end I will just have to keep getting more links, the more related the better.
Great tool! Link analysis can be so time consuming - comparing these charts with link velocity data should provide some really good insight into our competitors strategy.
Thanks for sharing, Tom!
Hi,
I'm pretty new to all this and we are embarking on our inhouse SEO campaign due to poor performance from our current SEO company ( after 3 years in position 3-5 for our main keywords we've dropped off to page 2 for all of them in one week ).
I keep being told they are in the process of buying more links. Does this article mean paid for links is bad or just some paid for links?
Could really do with some advice.
Great tool. Just tried it out and works well. It's a pain to export each time but never mind, nothing worth having is easy, and all that. Thanks again.
Tom - Great stuff! I've heard a fair amount of people (in Q&A) looking for tools that can profile their link spammyness and this looks pretty good.
When I get a spare moment, I'll definitely be testing it out.
PS. Nice dig on the 'math' front. Maths is by it's very nature plural and so that is how it should be written. Color also has a 'u' just before the 'r'.
So I ran this and we have two spikes. One in the 30 - 40 and a second in the 40 - 50. No links in the low neighborhoods. That is actually sort of scary. I do all of the SEO myself so maybe I am too picky as to where I seek leaks from? However, I do a fair amount of blog commenting and some of those sites don't have super authority.
Our correlation is 0.38. The others are 0.76, 0.69, and 0.50. We do have a fair amount of good rankings, so maybe I shouldn't be too worried.
Thoughts?
Yeah - I agree with some of the comments above. I don't think it is really, shall we say "identifying paid links" as alot of it is just pretty graphs however it does give you an indication.
Thanks for sharing!
Excellent tool, thanks for sharing. We hav our own model of this which is very similar but this is a bit easier to use so a time saver is always welcome!
I guess the one thing I think is lacking is any analysis on link text. Do you think this is important at this point or not? I've seen lots of link profiles which look pretty good on the surface but when you look and see 95% of all links have the same targeted link text then you can start to smell something fishy is going on.
Idea for version two perhaps? :)
Andy Adido
I am also having problems getting this to work.
I have added mt API and secret key to the config sheet as instructed but I'm getting errors reported when adding domains to the data sheet?
Wow. Thanks Tom!
Another great tool in the arsenal.
I really appreciate the generosity in this community and your contributions to it.
Tool Time ... :)
Sha
That's a great little tool Tom, thanks. Gives some fascinating results for the top 5 organic performers in my industry!
It looks to me like you guys are heavily involved in the large client / corporate scene. As a pretty raw beginner in trying to help my local business colleagues get found against longer established, more nationally orientated competition which has blanketed many niches / markets with geographically keyworded pages, (the SEO services market being the classic example), I am curious to know if any of you find the tool applicable in the less competitive (?) local geographical search work, where businesses are tryiong to tap into the search traffic in their own area.
Good question. I've not tried it with a business targeting a small geographical area. I have tried it with small businesses, but nothing like that.
You would maybe be better in those cases concentrating on looking at the top links from you competitors and their anchor text distributions. Good luck! :)
I am getting errors after the insertation of values. I am getting error under Normalised Data and Correlation Data. (#Name?). Any suggestions....
Tahir, keep the file in Google Docs and it will work alright. (And without the API codes if the free API)
Brilliant post Tom Anthony I used your discovered tool and it really works good for me.I am new to Moz and frankly speaking after joining my impact on the SEO is almost double thanks for the great posts shared over here the same as you did, Well thanks for the tool Tom.
Thanks for the really useful post Tom, it's great to be able to better visualize a site's link profile. I do have a question. Across many of my clients I'm seeing almost identical link profiles, could it be possible that this is due to the limitations of Open Site Explorer? Most of the domains I am looking at have more than 10,000 linking domains (I believe this is max amount of links OSE will report in CSV files). Do you believe it's possible that SEOmoz's method for selecting the best links to report, when there are more than 10,000 links, would cause them to select similar links for each domain, making the link profiles seem identical when in reality there's a lot of info being left out?
Perhaps, this is a question I should ask the SEOmoz engineers, I just wanted to see if you could shed any light on the situation! :)
Again, thanks for the wonderful post! The more ways we are asked to question and interpret data the smarter we become as an industry, great stuff!
Hey - great question! I reached out to one of Linkscape engineers to help get a detailed answer for you and here is what he provided. I hope it helps answer your question, but feel free to contact me if you want more answers!
"Our method for selecting links to keep when there are too many is to let the "least important" ones fall off the end. Least importance takes on one of a few meanings, depending on what view the user is looking at. For things sorted by DA, the smallest-DA links are dropped. For things sorted by linking root domain counts, the links with the smallest counts are dropped, etc.
Also, the number of links we keep for any given target (url, domain, etc) is 100 thousand, not ten thousand. In a few cases it is actually 750 thousand, but never less than 100k.
In regards to the question about the similarity of inlink counts by source DA, it boils down to what is meant by "similar". Organic link profiles have a similar shape called a "power log" distribution that kind of looks like this: a small handful of high DA source sites will tend to hold many of the links, and a much larger number of lower DA sites will each have a few.
The only variation I would expect to see is in the steepness of the curve. If someone were to focus energy on link building within only a small number of domains (perhaps because it is easy), their curve will look steep. If instead they only get one or two links from a wider variety of domains, it will be shallower. I personally believe organic link profiles tend to be on the shallow end."
Thanks!
Carin
We have been profiling links of players across a huge industry on study/volunteer abroad and I am excited to try out this shared technique. Thanks to you!
Tremendous infofmation likely special charts and graph show important information and easy to understand.
That is nice sharing information and i like it. Thanks for sharing informtion.
Thanks for the tool! In a lot of cases the Average Correlation is like animals herding together, making it easy to spot the outliers. But what might it suggest if there isn't a whole-ness of the herd? The range in the below Average Correlation graph is only .25, but it looks like a step ladder.
Graph hosted on Image Shack
I'd love to get everyone's thoughts on this...
Mike
Interesting. I find that one of my biggest competitor's gets about 49% of their links from DA 80 sites. It's heavily skewed, but in a direction that implies business relationships. I would guess that skews to the left to middle indicate poor quality link buying, while skews to the right are indicative of strong brands.
Pretty useful Tool, Tom. I have been running a lot of comparisons across several websites and have discovered a lot of new things. Not only that this tool helped me with competitor analysis, this tool has also helped me identifying the kinds of backlinks I was getting.
Kudos for sharing this!
I'm having trouble with the sheet. I put my correct API credentials in (pro). When I enter the domains in c4-g4 they cause an error in c5-g5. Error says a lot but toward the bottom it says "server response. invalid signature".
Can someone help?
Just gave the tool a shot and it worked well, per your instructions. However, something I'd like to point out is that for sites with a relatively lower number of links (like the one I tested) the distribution won't fit in nearly as well with the norms shown above. Guess that just means that I need to build more links to that site. ;)
Thanks for the insightful post, though! I love making SEO tools in Google Docs, and this is another good one.
You are right, with lower numbers of links it isn't as fine. That is a good time to edit the tool to pull in pages rather than root domains and sort by PA instead, maybe? Building more links is a good conclusion!
Glad you liked the tool. :)
Thanks for this tool! I compared my site to some competitors and it "proved" that I wasn't buying any links. I'm glad analyses like this can push blackhats out of the rankings.
Where is the proof that Google does look at the link profiles of other websites? Is there even any shred of evidence? Let's be logical here for a change. How can they understand who your competitors are? No, that's impossible. Okay then they might use the link profile of lets say Wikipedia - this is in another article. But wikipedia is one of the the most popular and linked to websites in the world so I very much doubt that has a 'normal' link profile. Lastly but not leastly, just the other day there was a nother post on Youmoz about how viagra sites rank. I bet they don't have 'normal' link graphs.
Unless someone actually does a study with some evidence then I can't believe any of this. It simply is illogical. Furthermore, if you don't pay for links, what will you do? Ask for them and surely that again will produce a link profile that is 'abnormal' especially if you are good at!!
However, as a general rule having links with a variety of text and from different sources is probably a good idea as there is a chance that Google might pick up on spam if you have the same anchor text over and over. However, even this is unproven. But it still might be good future proofing. However, the idea of a complex link graph is just stupid and frankly impossible. Sorry.
A complex link graph is impossible? Simply not true.
This is all a question of risk. You can go off and buy links, submit to hundreds of crappy directories, spin a load of articles and use Xrumer to dump a load of spammy link in forums. Doing that's time consuming, likely to cost you a fair amount of cash, and at some point - sooner or later - you'll get bitten.
The question you need to ask yourself is whether it's worth it. If you're a brand that trying to build a long-term presence, the answer is probably no.
The whole point of 'inbound marketing' for link building isn't that you go off and ask for links without giving anything back. Instead you're producing content that people find interesting and will want to link to. Infographics are a prime example. If you put a decent one together and market it to bloggers in the relevant niche, infographic sites, and seed it on a few social media channels then chances are (if you've done your job properly) you'll pick up varied links from different sources in a natural, and cheaper, way than if you simply went and bought them.
A complex link graph is impossible? Simply not true.
I never said it wasn't. What I said is there is no shred of evidence that Google compares complex link profiles and furthermore it is highly unlikely. Can you answer: How they would possibly know who your competitors are? And if they didn't look at your competitors, then what they would determine as a 'normal' link profile to compare you against?
Could Google tell that you are spammy in some way in simplistic terms? Yes. E.g. Position of links; yes we know that they started looking at signatures in forums. Variations in anchor text/alt tags; yes this is a real possibility but probably a minor factor only looking for basic things. Sites like Wikipedia and Wikitravel, for example, won't routinely have links that mention the domain name / 'brand' etc. so if Google starts penalising sites that don't have the domain/brand in the link (and it has been put forward that most 'genuine' links would mention the brand and go to the home page) then there is a big chance they could be shooting themselves in the foot for large informational sites. Just use logic rather than following the hype.
Is using a link graph a good idea to help you determine if you possibly look spammy - possibly.
However, the main premise of this article is unrealistic.
Hey Tom, This is pretty Awesome actually… now we actually have a tool to find out and conform that the link profile (if it is good or bad) instead of guessing around and say… “I think you like profile is Spammy J” so first of all thanks for that…
A while back there was a discussion around that following the competitor links can be harm full (as may be competitor have bad links in its profile) but I guess after this tool one can have a bit of a clear idea which competitor link profile to follow and what to leave…
Overall a great read specially a tool with FREE in it always inspires me ;)
Tom - this is so AWESOME!! A great help to visualize the data for the customers. Absolutely great!
its not helping find paid links this - its just showing link and their incumbent spikes... just pretty graphs really in my opinion, youll always gt spikes as you or the community link to you, nothing you can do about that other than devise a strategy to counter it, as for paid links - the only way to determin is to visit the linking site and see if they have a page about - "advertising" and so on..
Good read, but strugglin to see any use for this.
WITH THE HELP OF THIS POST, I'VE GOT SOME IDEAS REGARDING ANALYZING THE LINKS AND THEIR AUTHORITY VALUES. THANKS FOR THAT , VERY IMPRESSIVE!
Stop yelling, please.