We all know that links are good for SEO and good links are even better, but what does a “good” link profile really look like? It’s easy for even an average website to have hundreds of back-links, and sorting through them to get a sense of the overall quality is often more art than science. It’s also easy to get caught up in the outliers. Will 1 great link or 1 spammy link tip the balance? Probably not, but it’s easy to get distracted by those exceptions when you're filtering through hundreds or thousands of links.
Here on SEOmoz, we've tried to distill (and by "we" I mean a bunch of other people who are smarter than me) the idea of link profile quality into metrics like Domain Authority and Page Authority. These are incredibly useful concepts, but now we're on the opposite extreme – just one number to represent something very complex. The problem is, we really don't have much in between, a way to understand the quality of our link profile at a glance.
Link Profiling: The Experiment
This blog post really began when I wondered whether it would be possible to take our existing Moz metrics and chart what a link profile looks like. I went through a number of variations (subjecting Ben and Nick to harrowing emails loaded with dozens of graphs), until I finally landed on a process using Open Site Explorer. I'm going to outline that process, give a few examples, and then provide you a link to an Excel spreadsheet to download, so that you can play around with the idea yourself.
The basic process looks something like this:
- Enter a site into Open Site Explorer (OSE)
- Select “Show [Followed + 301]”
- Select “from [External Pages Only]”
- Select “to [All Pages on the Root Domain]”
- Export results to CSV/Excel
- Calculate the max Page Authority (PA) for each domain
- Sort max PA into buckets: 1-10, 11-20, etc.
- Graph the buckets
The result is a distribution of all of your linking domains by the highest-authority pages in those domains. This sounds a lot more complicated than it really is, so let’s see it in action.
Profile 1: High Quality
Let's start with what a high quality domain might look like – I'll make it easy and pick on SEOmoz. Using the process above, here's one way you might graph the SEOmoz link profile. Since Open Site Explorer exports a maximum of 10,000 links, I've restricted this profile to just the home-page:
You may be surprised to realize that, even for a high-authority site, most of the Page Authority is still on the lower-half of the spectrum. The simple reality is that even on a strong site, most of the actual pages that link to it are much weaker than their parent domains. Most of the back-linked Moz pages land in the second bucket, with a gradual drop-off as PA increases.
Profile 2: Medium Quality
Now, let's compare that with a solid but less authoritative site, my own blog. I've got solid back-links from some pretty good sources, but nothing like the Moz does. Here's what my PA profile looks like (this is also the data used in the spreadsheet below):
Here, you see that most of my back-linking pages are sitting in the 0-10 bucket, a clear sign of my inferiority (sniff, sniff), but the curve still levels off gradually and I've got some solid representation up the PA chain.
Profile 3: Low Quality
Finally, let’s pick on a site that came to us in Q&A with some trouble (we’ll keep it anonymous, of course). This isn’t a site that was heinously blackhat, just one that suffered from enough low-quality links that we suspected a problem:
Look closely, and you'll see a pronounced 0-10 bucket followed by a rapid drop-off, with little or no high-quality pages to take up the slack. It may seem like a subtle distinction at first, but look in the PA range of 20-70, and you'll see the difference.
The Excel Spreadsheet
You can download the spreadsheet (1.9 MB) and try it for yourself. Just export your own data from Open Site Explorer (as described above) and paste it into the first worksheet ("OSE Data"). The second sheet ("Domains") will automatically strip out the subdomains, and the third sheet ("Max PA") is a pivot table that calculates the maximum Page Authority for each subdomain and then collapses that into the 10 buckets.
One trick: You'll need to refresh the Pivot Table (how to do this varies a bit with your version of Excel). The other pages and the final graph should refresh themselves automatically. I haven't tested this on a Mac, so feel free to comment with helpful corrections.
This technique is a work in progress, and more of a way to explore your link profile than a hard analytical tactic at this point. If you try this out and find something interesting, please let us know in the comments. We're always looking for useful ways to enhance the data visualizations on the SEOmoz tools.
Update - DA Profiles
A few people asked in the comments about what the Domain Authority (DA) profiles looked like. Since the SEOmoz team re-normalized some of the DA data last week, these curves are very similar in shape, but I'm including them below (I've matched them on height of the top bar):
I really like the concept you've developed here, Pete. Part of me feels like this can be extended by adding in other metrics - anything from Linkscape's index or even third party sources. You could get PageRank from the toolbar checksum, get link counts from Yahoo!, get Compete.com rankings, etc. and build these cool graphical representations of what your link profile looks like.
I'd suspect that at client presentations and for in-house SEOs reporting back to their team, this kind of distribution and visual data presentation is awesome.
Quick question for folks reading this - is the graphical respresentations of this data something you'd want to see in Open Site Explorer itself? Is it of high value in comparison to other improvements we could/should be making? If so - any suggestions on how you'd prefer it was displayed?
I think the trick with these visualizations is crunching enough numbers to have a basis of comparison. It's easy to make a graph, but the "so what?" part is harder without knowing what "good" is. Many of our members probably don't realize just how much complex modeling and data-crunching goes on behind the scenes to put out even one useful metric.
Hmm... Great point - you almost want an "average distribution" overlaid on top of your own site's data and get those different overlays for a variety of types of sites (big media, government & education, e-commerce, social startups, etc.).
I really like this idea. The product roadmap here is quite full (as you know), but including this in the summer seems do-able.
I give a massive head nod to both of your lines of thought here. This is exactly the sort of stuff I want to see from SEOmoz both in the blog posts and in the tools. I'm looking forward to more of the same, and to seeing the visualization come to fruition in OSE. Great work guys, thanks.
Have been doing this and it's really interesting for clients. For it to be in OSE would allow some quick analysis.
What would be good is to have the keyword difficulty tool display the PA/DA chart for the full 25 that Excel exports, rather than just the current 10. :)
Pete, great article.
Rand, I would love to see this incorporated into OSE. The ability to view a side-by-side comparison of the link profiles of two domains would also be cool.
Also, do your statistics gurus have anyway to provide more insight and evidence of how a site's link profile impacts it's ability to rank. I wonder if it can be proven that, for example, controlling for all other variables 6 PA3 links are equal to 1 PA4 link, etc. This seems like a topic ripe for additional data study.
sistrix, a german SEO Service offers exactly that: a comparison of the pagerank of a set of backlinks compared to "the average". I am not aware what "the average" exactly is, but its a useful hint.
Another reason it would be great to have these charts actually in OSE is because many sites have more than 10,000 links. Since OSE starts with the highest authority to the lowest, it can skew these charts.
Hi Rand and Dr.Pete, both you "stole" my comment ;).
I was thinking the same while reading the post:
1) this can fast and make easier the analysis part of my job
2) this is a great way to help my clients to understand better what is a sometimes too 'heavy-to-digest" amount of metrics.
I'm going to experiment with you spreadsheet today itself.
TO Rand > I answer yes to your question.
The visualization should be a nice improvement to OSE. I do really like it and started using a lot both for deep analysis for acquired clients and short briefs for potential ones. And the main 'fault' is that the full list of metrics tab is not directly exportable in a file (you have to recreate it by yourself).
Sorry Rand... I didn't fully read your comment before posting mine below. Such visualization data in OSE would be awesome! Thanks.
I vote YES for seeing this kind of graph in OSE. Also linking domains by domain Mt in Linkscape.
I'm always for graphical representations. Even though they are easy to misinterpret - I still use them a lot, along with detailed explanation of what they mean.
yes...valuable in OSE
Hey Rand,
Do we want to see this in Site Explorer itself? YES! Maybe the more advanced version would be nice as you suggested.
Personally I would be helped with more comments/suggestions as in "what does this mean" or comments like a linkvoodoo produces. Still generic, but helpful.
My concern on a visual link profile is the accuracy of one single calculation: the PA. Have not been able to find how that PA is being calculated by SEOMoz.
Keep up the works guys!
Watch the blog tomorrow for more info on PA and DA :)
This is TREMENDOUSLY useful, and as much as I like Excel :-) it'd be phenomenal to have it in OSE.
Yes this would be great to see in OSE. To be able to show a client a visual representation of their link quality compared to their competitors would be very powerful.
Pete,
You lost me after:
I suck at excel. I have my spreadsheet open....can anyone walk me through step by step on how to accomplish the above 3 steps? (for some reason, it autoformatted to 1-2-3 instead of 6-7-8)
thanks
Grab the download and check out the 2nd and 3rd worksheet. I think it'll make more sense. You can just copy-and-paste in your OSE results and the spreadsheet will adjust (except for the pivot table). The formulas are all unlocked, so people can play around with them.
Excellent post -- it awakened me to the need to explore the Explorer at last! I have been playing with it today (and, parenthetically, gave it a rave in the SEOchat.com forums). I haven't gained full mastery of the tool yet but it completely blows me a way. SEO life has suddenly become easier. My favorite (free) link analysis tool used to be Yahoo Site Explorer in combo with SeoQuake, but I don't see myself using it that often any more: yours is a powerhouse by comparison. So easy to make clear sense of backlinks, and they do begin to add up to a "profile." Just filtering out this and that (e.g. nofollow) goes a huge distance. Heck, this beats by miles any link analysis tool I have ever used (or even dreamed of, to be honest). And by the way, I admire the newfound Software as a Service direction.
Thank you Rand and Dr. Pete!
That's a rave review - thank you :) Of course, I can't take all the credit (actually, any at all) for OSE. There are a lot of very smart people working to make that happen.
I tried.
And now I know my backlinks are full of rubbish..
thanks very much!
Happy to be of help(?)
oh god im too scared to run it for some clients... being how they have used several link building agencies in the past...
I guess it could also be used for evil and showing why a competitor ranks better because they have a good and wholesome backlink profile...
Dr Pete, that wasn't as bad as i thought, ive added mine to the SEOmoz facebook fanpage, if anyone wants to look.
That's awesome! Everyone can check it out here. I'd love to see more of these. Who else wants to upload theirs?
Too scared :)....
Really, really great post DP.Thanks for fleshing out your ideas with nice graphics that made it all instantly understandable.
This would make EXCELLENT added functionality to OSE and I totally vote for it to be incorporated.
Now I'm off to go play with your spreadsheet.
ok, im about to test this on my site, if im don't post a response its because ive fallen off my chair laughing...
stay tuned, maybe i should upload it to seomoz facebook fanpage?
I cannot wait to try this when I get home (the site is blocked at my government facility for some reason). I am on a mission to learn whatever I can on backlinks/SEO linking and I feel I'm far behind!
This has no relevance to anything, but I would just like to say that "Webmastermama" is my favorite handle in a long time.
Anybody able to get this to work in OpenOffice Calc? Max PA and Graph tabs don't seem to change when you paste the raw data for OSE into the OSE Data tab. I did 1-5 in the steps outlined then imported into the provided excel sheet (in OpenOffice).
Thx much.
What I find interesting is what a normal distribution should generally look like.
Having picked up a high profile website that has had poor link building, i.e. lots of high PA loaded, low DA out of context links, we are working hard to build links to camouflage the previous SEO agency's link building.
It's not just the PA too that should be profiled. Take the anchor text used to point to your pages, and build your own anchor text distributions across your top 10,000 links. We are also working hard to build non-hero terms that will reduce the disparity between the vanity terms and longer variations of it.
In my experience keeping PA/DA/anchor text distributions looking natural has always been critical to achieving success.
It's pretty easy to adjust the spreadsheet to profile Domain Authority (DA), but since the data changes last week, DA has normalized quite a bit. What you end up with is much more similar curves with varying steepness from the top. It didn't end up being quite as interesting.
Presenting this INCREDIBLE data in the form of graphs is HUGELY visual and I think it works really well for many scenarios.
If you're a consultant working for a client, you want to present everything and anything that helps deliver your message. If you're working in a team and you need to present your findings, graphs are great. You're giving a presentation to a prospective client, graphs work. Or even, you're a single SEOer needing to figure out what all this data, that you've collected, means and decide what to do with it.
Graphs present a visual perspective, even physical, to metrics which many would find overwhelming, particularly myself.
pretty useful, thx dr. pete !!
with xls 2008 for mac I had to redo the pivot table which wasn't refreshing
I think it would be uber useful for OSE to add a filter to show only the top page for each domain and not 25 pages per domain as it is today.
In many sites you might end up seeing only a 2/3 domains for the first 50 links u have
I'm also on mac but no really into Excel how you redo the Pivot Table... if you know on Excel 2008 as it's not see as a valid pivot table for now :/
Maybe it is just me. On my version of Excel 2003 I do not have the ability to refresh the pivot table. I need to edit it on a later version of Excel.
In terms of distribution of links, it got me thinking about link buying. Maybe one of the indicators to Google is an abnormal distribution of links. One expects links to see a bunch of high quality links and a bunch of low quality links, not skewed completely towards high quality.
Of course. Especially uneven anchor text distributions can trigger an automatic filter.
Hmmm... sorry, I'll have to see if I have the older version on another machine. I saved it in the more universal format (not ".xlsx"), but I wasn't sure what functions might have been deprecated.
The pivot table doesn't seem to work in NeoOffice under OSX. NeoOffice Calc has something called "data pilot" that is reportedly the same feature as Excel's pivot tables, but when I refresh the sheet, the resulting graph shows nothing in the first two buckets.
Great application for the Open Site Explorer, and I want to give 2x the thumbs up for the handy Excel sheet. MS Excel is my first love, and I always like to see new ways to crunch formulas and analyze data.
thanks for the post, the idea is really useful for many SEOs
Really useful visualization of the OSE data, really interesting when you look at the link-graph morph over time on a site you have been working on the past few months.
Wow, this is great. I had not fully explored the Open Site Explorer but this data is great. There are still a few things I like about MajesticSEO.com though that I don't see in the Open Site Explorer. Such as listing the number of domains linking to you (or maybe I just missed that when looking at OSE). Also MajesticSEO seems to have a larger index of link data. But still OSE is very useful! Thanks for the great post!
Amazing post! I'll give another vote for seeing something like this incorporated into OSE.
Awesome post! I do have a question though, on step 6 you said "Calculate the max Page Authority (PA) for each domain" what exactly does that mean? Does that mean taking the DA and the PA and averaging it for the page linking to your domain?
Sorry, I didn't make the connection between those steps and the attached spreadsheet very clear, but there's a worksheet that performs that calculation. Basically, it's two steps:
(1) Strip all of the linking URLs down to the root subdomain.
(2) Collapse each subdomain down to 1 value, the link in that subdomain with the highest PA. I used a pivot table for this.
Thanks good doctor!
My domain has the same authority as yours, i.e 64. Seeing as I expect your domain to have a good authority score this makes me think I should perhaps be monetising it better or at least making our site more attractive to linkers/advertisers :-/
Thanks, this is very useful even today, after 2 years.
Why SEOmoz has not created a similar tool online?
Salvino, maybe it's because we had a ready made tool for it?
https://www.linkresearchtools.com/tools/blp/
https://www.linkresearchtools.com/news/link-profiling/
Best, Christoph
Hi Guys
Just joined the SEOmoz community and love this article. However the Pivot Table will not refresh to include my domains - it still lists the domains within the original OSE Data tab. I have used Excel 2007 to try and "refresh" the tab but it only gives me options to change the sort order.
How are the domains in the MAX PA tab generated? I can't see any formula on the cells.
For info, the Domains tab updates no problem with my URL's from OSE Data.Thanks again.Kris
As soon as I posted this I managed to fix it. Why does that happen!?
Thanks for a useful tool guys.
Hi Pete,
i made the exercise with your spreadasheet and fresh OSE Data, for all the domain of a travel site about denmark, only followed links and 301.
Here is the graph: travelsitedenmark.jpg
What is your diagnostic, please?
Thanks,
So, first off, let me say that this is intended as an exploratory tool - it's a glimpse into your profile. It is not a full-fledged diagnostic, and no one should simply look at a 10-bar graph and claim they understand your link profile. Also, PA is just one measure of link value. Rand has a great post on the subject of link value that I highly recommend.
My gut reaction is that you've got a clear shift into the second "bucket", and that's generally a good thing. You do have substantial presence in the first (1-10) bucket, though, and I would dig deeper into that. It could just be relevant links on long-tail pages that happen to have low PA, or it could be indicative of overusing low-value tactics. There's no good way to tell without digging into the links themselves, but that's where I'd start.
Hello Pete,
Thanks for your valuable feedback. Is there a way in OSE to filter out the links only with the first bucket, or can I do this only in Excel?
Interesting is that I have another site running with much less Link Rooting Domains (664) and with much less Domains in the first bucket. For this second domain i do not have much used low-value tactics and not it is outperforming right now the competition at Top1 for the most competitive keywords in our segment.
travelsitedenmark2.jpg
Second the speed of link building was the lowest of all my projects and now the best performing.
The first one has 1006 Link Rooting Domains. My strategy for the first domain is now trying to acquire Links with high PA and DA.
I will do now the metrics using your Excel-Sheet with a variation to check the DA distribution: johnfdoherty.com/DA-TEMPLATE.xls
Why you do not an interface for this kind of visualization in OSE or SeoMoz campaign level?
In any case we have serious On Page problems since the upgrade of the first site at the end of last year, so I am not sure, if these link factors influence so much the ranking. Since Penguin, the site have lost only a very little in the search visibility index, but the second have gain search visibility.
Any way the quality of links is now much more important than the quantity of links.
This is my resume of: After Penguin, What to do.
Greetings from Germany,
Unfortunately, there isn't currently a way to filter. Exporting the full data to Excel is really the best bet. We've been exploring some options for visualizations, but a lot of it comes down to computational issues - what we can individually doing in Excel is a lot tougher at very large-scale in a real-time interface.
Okay,
in the past I work several years transforming xml data with xslt stylesheets, for a financial software company as software developer. The difficulty is to have the data, if you have the data like you, visualization, order, sort and filtering is quite more easy. Do you have seen the last visualization from your competition MajesticSeo, it is quite cool: https://blog.majesticseo.com/development/flow-metrics/
Dr. Pete,
After reading through your post and following your process, I realized that I'm missing a lot of data just being a Free member. This makes me want to go PRO sooner :) I tried refreshing the pivot table on my Mac on Office 2007 but unfortunately the data won't refresh (the Refresh Data option is greyed out). I tried it on my PC and it works fine.
Look forward to reading another one of your posts. Keep up the great work!
Jackson
Every time I convince someone to go PRO, an angel gets its wings ;) Seriously, glad you found it useful.
I've got a site with almost half a million links. How can I export this data out of OSE despite the 10 000 row limit?
Great post.
In OSE, how come you have a tag on a social bookmark site "https://www.bookmark4you.com/tag" generating a PA of 23.
If a simple tag is PA 23, it suggests that most of SEOMoz's links are worse than a simple bookmark tag. (According to your graph)?
Surely this shows that by having lots of "tags" you can get your quota of the PA 0-30 category links?
FINALLY,
Is it ok to just focus on high quality PA links as a strategy, or do you need the low ones as well?
Keep in mind that we re-calibrate PA/DA over time as the model improves, so it's hard to compare a current site to an older post, like this one. It's also worth nothing that PA is like PageRank - it's non-linear. Getting up to the 20s is pretty easy, but then it gets a lot harder from there on out. Each +10 is progressively tougher to achieve.
It would be incredibly rare to have a link profile that was only high PA links, unless they were artificially inflated (in which case, PA may not reflect their real value). I don't think you have to worry about going too authoritative - you'll always have some lower-authority links in the mix. If you somehow looked at your profile, and it was skewed toward the top, that would be a nice problem to have.
This is pretty useful and would be good to see as part of the online tools. One question: I checked 3 sites and there always seems to be a row in the "Max PA" sheet near the top where the URL field is "#VALUE!" and the "Max of PA" is 0. Anyone else getting this? Is this an error?
Will certainly try this out.
Is it actually working at the moment, as I'm building links for a website but Open Site Explorer is not showing any increase in links. My client isn't too happy although I can show them every link I've created.
Any ideas?
Hi Pete,
thanks very much for this awesome and inspiring post!
We've been doing link profiling for a while for a couple metrics, including ACrank, MozRank, Pagerank etc.
I'm happy to say that we will fully automate the link profiling in the same fashion as you describe in our TOMORROWs release of our "Link Research Tools" at https://www.linkresearchtools.com
I would be very glad to get your feedback on it once it's out and also be grateful to provide a test account for you, after all we're aggregating SEOmoz linkscape data just as we do with Majestic, Sistrix and the "public" link data from Google, Yahoo etc.
Just let me know!
Cheers,
Christoph
As promised yesterday,
we now have link profiling built in - fully automated ...
you can read up on it at
https://www.linkresearchtools.com/news/link-profiling-a4uexpo/Looking forward to your feedback!
Thanks - I am off to run our report this minute!
Nice Dr. Pete, looking forward to finding some time to play with this. Would be awesome to see some of this rolled into the Linkscape reporting.
I'm not sure how to get the pretty bar graph Dr. Pete shows us, but just sorting the sheet by PA low to high and then using a line graph it's obvious that most of my test client's links are in the 10-20 PA range. What are the actual steps in Excel? I can sort them into the "1-10" buckets quickly by multiplying by 10% and setting the rounding to zero decimals, but then I'm stuck on how to go from that to the bars. Or is the simple line graph accomplishing the same thing with fewer steps, even if it's less pretty?
There's an Excel download at the end of the post that takes you into some of the details. You can just cut-and-paste your OSE results into the first sheet and then the other sheets do most of the work.
Part of my goal was just to collapse the data a bit for clarity, but part of it was to collapse across domain. If you just graph the raw data, you end up with multiple entries for each domain (up to 25 with OSE), which can skew the results a bit. The difference won't be dramatic, in most cases, but I wanted to make the data as clean as possible.
Ahh, yes, unobservant me. Thanks! I'm guessing for my clients, who are rather small, the difference may indeed end up being significant since those multiple domains would comprise a larger proportion of the total link picture.
Yeah, if your client has, for example, 300 back-links and 200 all come from the same directory, it'll definitely make a difference. I see that all the time. Personally, I think it's always valuable to look at data in different ways. Any method of visualization is missing some angle, and if we stop and think we've found the "perfect" graph and only use that one from then on, we'll eventually make a huge mistake.
#1) Great post, elegant solution. Anytime we can present clients with visualized data, the easier it is to convey a concept
#2) Rand recently responded to my Q & A regarding feedburner links in open site explorer. OSE is a great tool, but is only as smart as the person using it (or rather, the person who answers the users question ;). I think its one of the best link profiling tools I've used, and there is still a need for educated insight and interpretation.
Pete, Rand, anybody - how can I/we learn more about OSE so as to better understand and interpret the data?
Thanks for reminding how useful is Q&A.
Just because of it is due to be PRO.
Thanks Dr Pete, now it's even harder to resist getting a pro membership (argh!)
resistance is futile (it had to be said)
Neat idea Pete and nice visualisations. This is something we've been doing as well but hadn't considered the max PA metric... I like it!
I am very weak in calculation I expended 15 minutes here also down load the excel sheet I may not get the points but no worry one of my favorite SEO Rand Fish here I am sure he will come again with easy computation, Now I am still working with my content ha……….
An interesting way to visualize link profiles. Simple yet clever. Nice job.
It would be interesting to produce such a graph monthly and stack the values to help visualize link growth over time - especially if put side-by-side with competitor's data.
I've been looking for more data analysis ideas for my websites and I believe this is the ticket. Now that I know what I can analyse I'll a lil bit wary of doing it. I'm pretty sure it's going to be quite bad! I'll do this when I get home.
Thanks for the suggestion on link profiling Dr Pete!
This is fantastic!
My question is, how trusted is the end analysis, if say, your site has 5x the links that you are able to pull into a csv? Since you are only pulling in the top 10k links, and not including the "bottom" 40k, wouldn't this skew your chart toward a link profile of better quality, since you are looking at 1/5 of the data? Does it matter if we are looking at top 10k, vs. top 50k, vs. top 500k, or top mil.?
It's a very good question and something that came up in my own data collection. Honestly, I'm not 100% sure if there's a specific order by which 10,000 the links get sorted, so it's possible the data could be skewed. For SEOmoz, I restricted it to just links to the home-page, but that was still over 10K. I'll see if I can get a better answer on that from the Moz data wizards.
Thanks! I compared root vs. sub vs. page of the same domain and the output chart was different every time, it's a great comparison tool to see link profile quality across these metrics.
Apparently, the first 10,000 links are sorted by quality, so if the data is cut off, you're getting the top links. This is done by Domain Authority, though, so the sampling of Page Authority should still span the spectrum. It's definitely an issue to consider, though.
Thanks for looking into it!
FYI - Your spreadsheet link doesn't seem to work
This web page is not available.
The webpage at https://seomoz.box.net/shared/r5762rfltp might be temporarily down or it may have moved permanently to a new web address.
More information on this error
Thanks for the heads up, Ben - checking into that. Might just be temporary. The spreadsheet ended up being a bit larger than I originally expected (it includes sample data from my own site).
Double checked and it works (at least for me)
hmmm bizarre. May be something to do with the Proxy im sat behind at work.
Love, love, love this article. Those graphs will really make it easier to get my point across when I'm trying to illustrate where a site's strong/weak points are.
Thanks Dr. Pete!
Very similar to the charts www.linkdiagnosis.com shows, only those provide pagerank distribution.
Very useful for fast comparison.
Thanks - I just checked that out. The data is definitely interesting.
Note to readers: The tool is Firefox only and needs a plug-in (just to save you from trying it on all 3 browsers like I did :) ).
Only just read this properly Pete. I like a lot. I tried to do something like this a while ago ("advanced link charts" or some such) but yours is better. Good work (this time).
:)
My link profile tells me that most of my links are not "bad", but that I should try to find more authorative websites to link back to me (PA > 30).
I wonder if the fact I'm linking principally from french websites influence the PA of my links. I mean; from an international point of view, it's easier to find authorative websites in english, since it's the "Web Language"...
Hint for ppl in the same situation as me (Website in foreign language): Having an english version of your website can be a big help in order to find authoritative backlinks!
Pete, it gives me an idea for a fully automated tools, using Linkscape API, to produce this kind of chart in a mouse click... Could it be interesting? :) I've been looking for a way to test Linkscape API for a long time...
Anyway, thanks for this great post! It's really awesome to be able to see and understand data in a visual way like this!
The team is exploring adding this type of visualization to OSE, but the API may need some development first. The export can handle 10,000 links, but crunching that data via the API is tricky and resource-consuming. That's why I wanted to put the idea out to the community first, and try to refine it a bit. Happy to see the positive response.
What you say about PA for not english tongued website is something I suspect it's true, as it's something I see also checking out very important italian sites.
And I suspect that this factor influences also the keyword difficulty tool, as I find a little bit strange that very generic keywords in Italian are set as "no competitive". Tell me if you have the same opinion about this doubt I have.
Finally, the tip to have an "international version" of your own site is something I'm considering, also to expand the link adquisition on a more wider panorama and to start guest blogging on my site.
Honestly, according to the tests I've performed, the keyword difficulty tool seems to give results that are totally relevant.
Take, for example, the keyword "posizionamento" (Which I took from your website, since I don't speak italian).
Keyword difficulty tool gives me this:
posizionamento - Google.it - Difficulty: 43%
I compared this result with its french equivalent, which is "positionnement". It gave me:
positionnement - Google.ca - Difficulty: 43% (I'm a french canadian)
positionnement - Google.fr - Difficulty: 43% (For my european cousins)
Please note that even if difficulty percentage are the same above for Canada and France, results are different (apart from Wikipedia being the top result, which I'm aware can play a role).
You can note that it produces exactly the same result for this keyword, even if the language isn't the same.
Then, I run it in english (US):
positionning - Google.com - Difficulty: 54%
Boom! +11% difficulty. How can we explain that?
Well, my previous statement "Having an english version of your website can be a big help in order to find authoritative backlinks", is not reinventing the wheel. People from around the globe already understand this reality, and they got an english version of their website, even if they're aiming their local market - It gives them a potential for backlinks, but mainly for international business.
The result of this is simply that it brings more competition for a same keyword if it's in english.
Look at this simple example:
Let's say an american, an italian and a frenchie want to start a business.
Each of them have a website which targets the generic keyword "positionning.
The american has an english version of his website, which is obvious. He's using the keyword "positionning".
Our italian guy has an italian and an english version of his website. He's using both "posizionamento" and "positionning".
Frenchie has a french and an english version of his website. He's using "positionnement" and "positionning" as his keywords.
For this simple example, which involve only 3 websites, you can easily see that english keywords get the most propensity: "positionning" being used 3 times, while "posizionamento" and "positionnement" being used just once.
In conclusion, I think this is the main reason why english keywords are more competitive than those in foreign language... It's because we keep using the english language as the standard for the Web! And if we want our website to be seen by everyone on the planet, the easiest way is to have an english version of it. It gives you plenty of benefits, the first being backlinks attraction, as well as many others I won't describe here because my comment is already really long lol.
Maybe I should wrap up a YOUmoz on this... ;)
Tell me if you agree...
Hey Ptech. Thumbs up dude and I agree. I'd love to see a YOUmoz post on it. Get to it!
I agree totally with you, but that doesn't mean that some results are making me thinking.
As I agree completely with your "italian guy" (as I did the same searches about my site), I can tell you that all SEO's keywords are a world apart, as object of the desire of all SEO people.
I'm talking about other market keywords.
For instance "aria condizionata" (air conditioning) is a very common term, but keywords tool indicates a 16% of difficulty.
The same for "passeggini" (strollers), with is only 35% (moderately difficult)...
So, the question is: does the tools needs some adjustments or I've been so lucky to find "gold mines"?
One thing that can be said and that can influence those % is that usually in the first rankings we can find pages from Ciao, Kelkoo and other webs of the kind, which have usually a poor PA but a great DA... and the DA is the metric to look at in these cases IMHO.
I understand your point, but I don't want to commit myself in saying that DA should be considered over PA in this case. I explain myself...
I made a simple test with your "aria condizionata"on google.it, and the first website that comes up (at least on my side) is publiweb, with PA:0 and DA: 50.
2nd is ariabox, which seems to be more related at first glance, with PA:35 and DA:51.
When you look closer at these results, even if the 2nd result got better metrics, it's still in second rank:
Position | PA | DA
1st | 00 | 50
2nd | 35 | 51
After analyzing both of these websites, you can easily find out that publiweb has better OnSite optimization than ariabox.
Now my question would be; Is the keyword difficulty tool taking OnSite optimization in consideration as it should be? My guess is that it's not. It's mainly building up a link profile for a given keyword, and returns an indication on how it seems to be competitive, according to this profile.
As Pete said, I think that the Keyword Difficulty Tool involves the same concept described in this post; It's more a "way to explore [a keyword potential] than a hard analytical tactic".
You could have a keyword that has a difficulty of 6%, but for which 3 extra-strong websites are fighting. Does it makes it more difficult than a global 6%? On my side, the answer is "yes".
Does DA should be considered over PA? In most cases, according to my personal tests, no! But if some ratio could be made between PA and DA, I think it could be even more appropriate.
Again, this is just my opinion. Thanks for sharing yours!
Thumbs up from my side.
What you say is substantially correct. And helps me in using better the metrics of keywords difficulty tool.
But - and I'm sorry I can't give you real data (watching/listening to football Inter vs. Barcelona right now) - following our "brainstorming", I still think that DA is something to be considered.
Let's say that we have two websites both greatly on site optimized and with poor PA. In that case, and that should be I would likely propose to the weaker one on the rankings side to create a link building campaign in order to obtain:
1st) a greater PA for the category products page using as anchor text the main keyword "aria condizionata" and semantics
2nd) and improved DA in order to give more substantial power to the category page too through a well organized internal linking.
In this sense the keyword tool can be used as an instrument to plan your SEO long term strategy. This was also the meaning of "gold mine keyword".
And I agree with you and Dr.Pete: the KDT is a way to explore possibilities, but if used with all the others tools and experience it can help to design the better option for your strategy.
Love our chat! I hope it is useful to all the others here.
Yup. I totally agree with you! I would go for the 2nd website too (ariabox), even if publiweb is ranking 1st.
I also enjoy our chat! All of this to say one thing in the end: KDT data is subjective, and as you said, we should focus on the big picture when selecting potential websites to get backlink, according to our SEO feeling of what it represents.
This is a really pheonmenal discussion, but it's also convinced me that I need to do a blog post specifically on keyword competition/difficulty. Hopefully I can put that out early next week so we get into more discussion on this.
And just FYI - I think all your instincts are spot on (english = more competitive usually and PA/DA don't look at on-page optimization, nor does the KW diff tool). The PRO Webinar (now archived) on reverse engineering the SERPs is a really good one for this topic, too.
"The power of community".
Thanks Rand... and a promise is a promise ;)
I'll check again more quietly the webinar.
*group hug*
lol...
Thanks Rand. I'll be looking forward for your post next week! :)
Rand, you're going to have to update your pie chart to include Page Speed. :) Any ideas on % for PS?
Great discussion! I was looking into this part of SEO (international) and haven't been able to land on something concrete until now. I agree that site, which are in foreign languages, should have an English version.
Have we taken into account the search engine language? For example, French Google vs. English Google. Depending on which language of search engine the user is using, IP, and keyword? that the results will show up differently. Am I correct/wrong on that?
Vim thank seen in my blog. And say that I liked the blog.
Thanks for this tip, great help.
Can you add a chart for DA so we can compare?
I did and it looks just like a bell curve.
https://yfrog.com/evpadaj
I would imagine the bell curve look for DA is going to be the norm.
You're essentially correct. After the renormalization of the data last week, the DA curves have similar shapes, varying in the size of the peak. I ran a few variations before picking this one, but I'll try to dig up the DA graphs today and append something to the post.
The DA graphs look like they're flipping us all off... just sayin.
haha totally
Hi there,
Loving this concept, however I opened the spreadsheet and cannot see how to calculate the Max PA. Can someone walk me through this bit please?
Much appreciated.
Unfortunately, it depends a lot on your version of Excel. It's a pivot table, and it has to be manually refreshed. In my version, there's a "Pivot Table Tools" icon at the top, and then you select "Options" and "Refresh".
I tried to make it as hassle-free as possible, but I couldn't figure out how to get around the pivot table. If any Excel wizards have an alternative, I'd welcome it (and would re-post that version).
Hi ok,
Thanks for taking time to explain that to me. I refreshed it and now it works for my data. Thank you.
Is there any more resources that I can read to learn more about page authority and maximum page authority?
Good work by the way.
Cheers!
I'll give it go and report back, it all sounds like useful stuff!
Thanks for all your thoughts and inputs..top post
Chris
(Link removed)
Dear Chris... just a suggestion for your thumbs up health: do not link to your own site in your comments because
1) people here will bomb you with thumbs down;
2) the link is going to disappear in few hours, so it's not going to give you any real advantage.
More about the policy of the SEOmoz/YOUmoz Mozpoints here.
Thanks for jumping in and helping out!
point taken..sorry