|Link development risk assessment|
I'm putting together a presentation right now that's caused me to clarify some of my link building. One of the things I'm trying to pound through their heads is the idea of risk assessment. Not coming out and saying go one way or the other, but simply being able to assess the risk so a decision can be made. Here's the idea:
- you must measure three things about the backlinks: authority, relevance, and footprint.
- I measure authority by backlinks - if a site has links from authority sites, then it is an authority.
- I measure relevance from content and backlinks.
Links from sites that are neither authoritative or relevant can increase risk.
- footprints. If something is mechanized, then it is likely to leave a footprint. Just like everyone is trying to reverse engineer Google, Google's reverse engineering your methodology. And Google's smarter than you. they're also better looking. And have more money.
Examples of footprints:
- type of site (i.e. all blogs)
- all low end sites
- look at link attributes for footprints: position on page, anchor text, incontent, there's more.
- one thing that is often overlooked, network footprints. Anytime a closed set of sites is used, a network footprint can be found. Even if the set of sites is large and diverse.
Any time a system is used to generate links (and a system is used almost 100% of the time), that system will leave a footprint of some kind.
Let's say you crawl the web looking for sites that match a specific criteria. You get a weekly list of potential sites and contact them for links. All great, relevant authority sites. No footprint right?
Oh yes there is - all the sites you get backlinks will meet the original criteria from the crawler were sites were selected to be contacted. That's a footprint. And it's probably a footprint a mile wide too.
Assessing the risk then comes in to play based on whether you think Google measures this, as well as if you think Google will measure this in the future.
Good idea. We use a pre-qualification list after prospecting.
|- look at link attributes for footprints: position on page, anchor text, incontent, there's more. |
Relationship with other websites, types of websites previously advertised, text on the advertising page referring to how the website advertises and number of pages without advertising versus number of pages with advertising (aka outbound text links) is important in my experience.
|Oh yes there is - all the sites you get backlinks will meet the original criteria from the crawler were sites were selected to be contacted. That's a footprint. And it's probably a footprint a mile wide too. |
Absolutely. On way around this is to create a link grid, where you pursue links on a priority level based on a random series of BOTH website type and anchors used.
So for example, you might do the following:
Blog / guest post
organization / association link
resources page link
top 10 blog post listing
link bait inbound link
Blog / guest post
This throws diversity into the chain of events further reducing the pattern. (Inevitably there are only so many ways to develop links, that some will *always* be patterns)
Map out anchors according to competitive aggregate anchor levels on a sliding scale from aggressive to low-threshold, and map those directly into an excel or similar chart so that you can stay diversified.
Stay away from networks altogether is my experience.
|Inevitably there are only so many ways to develop links, that some will *always* be patterns |
I'm not sure this is a bad thing. In fact, I think too much diversity would be considered unnatural. Actually, I think even a moderate amount of diversity would be unnatural.
Most sites nowadays (unless they are a mega news site of some kind) are going to appeal to a particular demographic. And I think those in that demographic are going to have a preference for a particular linking pattern. Each demographic of link partners is going to have certain linking tendencies.
Some sites have content that is quite topical and less evergreen, so why shouldn't the lion's share of links be from blog posts and comments (and tweets)?
Some sites will appeal primarily to people whose main interaction on the web is through facebook. So why shouldn't those sites have overwhelmingly high amount of facebook likes and mentions?
Some sites are scholarly, and you would expect university professors to link to them, but you wouldn't expect facebook likes or tweets or other links to them. You would probably expect very few links from blog articles to them, since blogs are overwhelmingly topical and those scholarly type posts would be more evergreen in nature.
I bet that if you analyze anchor text, there would be a certain LACK of diversity based on who is doing the linking, too. For example, I bet that professors probably use the exact title of the destination page as anchor text more so than bloggers. The way professor use anchor text is PROBABLY influenced heavily by the way that printed publications use citations.
Bloggers probably would be more likely to use the writer's name / publication's name somewhere in the anchor text, I am guessing. Something that is more headline grabbing rather than formal. In fact, the call to action that is often found in anchor text (such as the words "click here" in the anchor text "click here to view wheel's latest article on link building") would probably vary depending on whether it was written by a professor or a blogger.
I'll bet that "hobbyist" webmasters are more likely to link to sites using the destination site's NAME as the anchor text than other types of webmasters would.
Along those lines, I bet if you compared the linking patterns of blogs to scholar sites, you would notice predictable differences between the position of the link in relation to the surrounding text. I be that a professor would be likely to put the link at the top of bottom of the surrounding text than a blogger would. Conversely, a blogger would be more likely to put the link in the middle of a paragraph rather than as a "stand alone" link at the beginning or end of a paragraph.
Does google have the time to worry about all this? I don't know. I am sure they are certainly CAPABLE of profiling sites and could easily spot a link that stands to far outside the typical linking patterns. I just don't know if they are currently looking at it that way.
Just my two centavos...
Great post planet13. :)
|Some sites are scholarly, and you would expect university professors to link to them... |
Shaking things up a little... Is it possible that dot edu links aren't such a good thing for a commercial site, IF a dot edu sends Google the signal that your site fits into the Scholarly Query Bucket and not the Commercial Query Bucket?
Wikipedia has no problems ranking on commercial terms, at least not in my market
|Bloggers probably would be more likely to use the writer's name / publication's name somewhere in the anchor text |
Try it. You'll likely find that left to their own devices, they'll just use your company name.
Point is, as soon as you use an artificial system, there are recognizable patterns and you've reduced diversity. This increases risk of a Google penalty, because it can be found.
For example, remember when some level of anchor text was a good thing? This year, it's a bad thing, anchor text pointing at the page that's doing the link building is the new good thing.
If you're using any kind of system, adding a random factor doesn't remove the underlying pattern. And if/when Google decides to look at that underlying pattern and devalue it while perhaps inflating the value of some other factor, that's when you run into problems.
I watched a presentation at pubcon where someone had a PhD in statistics tear down and rebuild the link profile of ranking sites - they know the factors that lead to rankings. Then it's just a matter of duplicating those factors so that your backlink profile matches what's ranking. this isn't that difficult if you have the data and can guess at the probable factors (which can then be tested to see if they match). Everyone sat with rapt attention listening to this. Sounds great - and modelling a system is well known technology. The only problem? In Google's case they are prone to changing the underlying model on you, so your sites no longer match the system or the model. Panda's a good example of changing the model rather than just the variables.
It's all risk. If you look at your link development technology and believe that it's low risk, that's fine. But don't imagine it's 0 risk.
In the context of my presentation (end users) the problem is that EVERY SINGLE SEO company out there is going to tell them that their system is white hat and 0 risk when in fact every single one of them has some risk. The better SEO companies likely tell their clients that there's hardly any risk, but disavow any responsibility. In the presentation I mentioned above, no mention was made of what happens to your site when the underlying model changes (and that's from a company I label as 'genius level'). In any event, end users are not being appraised of the underylying risks. I'm doing so with these clients, I want them wary when they hire an SEO company - that's really the attitude one needs to take.
|Wikipedia has no problems ranking on commercial terms |
Depends on how you define a commercial term. Is fishing lures a commercial term? It has a commercial component to it, but it's not a commercial term. This is why wikipedia ranks for that phrase.
Cheap Fishing Lures is a commercial term. The word cheap is an explicit clue that the query is commercial in nature because it is about price.
I am not doubting that you are referencing a true commercial term. I am simply stating that we sometimes think certain phrases are commercial when in fact they are not. Yes, I agree that a certain percentage of those making these queries have commercial intent, but there are many others who do not have that intent. That is when you are going to see wikipedia rank and why you see it ranking there.
|Is fishing lures a commercial term? |
It's sure as heck a strong buy signal when you throw that term in front of me :).
|If you're using any kind of system, adding a random factor doesn't remove the underlying pattern. And if/when Google decides to look at that underlying pattern and devalue it while perhaps inflating the value of some other factor, that's when you run into problems. |
No, but it reduces the risk by one more level. Any more toward diversification of the link profile build reduces risk.
Definitely, there is always risk.
Then there is the law of diminishing returns. The further one works to reduce the last percentage points of risk, exponentially does the workload increase.
I would say find a level that you are comfortable with - and one that ultimately is much more thought out than competitors - and you are probably doing ok.
|Shaking things up a little... Is it possible that dot edu links aren't such a good thing for a commercial site, IF a dot edu sends Google the signal that your site fits into the Scholarly Query Bucket and not the Commercial Query Bucket? |
Well, from where I am looking at things, that seems to be the case...
Panda was GREAT for my ecommerce site; my site is up 30% since pre-Panda days. I think some .edu links (as well as non .edu links from some information-based, non-commercial sites helped out here significantly.
But sales are down significantly. Lots more traffic = far fewer shoppers. Oh well...
Starting to wonder whether having information-based articles on an ecommerce site (and links from other information-based sites) might not be a liability.
|You'll likely find that left to their [blogger's] own devices, they'll just use your company name. |
We don't have many bloggers linking to us at all from IN CONTENT links, but when they do, they tend to use a kind of "tease" about the article. Our own names aren't prominent (although the business name is), so that's why in our case they probably don't use our own names.
Of course, blogrolls will use the name of the blog, but that's a whole 'nother kettle of fish.
The "hobbyist" sites overwhelmingly use our company name (again, as far as I can tell).
|Point is, as soon as you use an artificial system, there are recognizable patterns and you've reduced diversity. |
I agree with that. But I am simply wondering whether backlink "over diversity" might not be an easily recognized pattern by google. It certainly is an "artificial system." If the majority of sites have X amount of diversity in their backlink profile and we have 3X amount of diversity, I wonder if we might not be placing a target on our backs...
I guess when I hear things like, "We only count the links we trust," I can't help but wonder if they actually still "trust" a link from an .edu page to an ecommerce page as much as they did before. Maybe JC Penny left a bad taste in google's mouth...
Sometimes I wonder if that is part of the reason that big brands are doing so well in the SERPs lately. Yeah, they have a massive diversity of links, but they have ALWAYS had a massive amount of backlink diversity.
But suppose for a moment that google was going to more highly value commerce-related links to ecommerce sites. Those big brands probably have a higher percentage of links that are on some public forums that are related to buying something, like, "Where's the best place to buy an SEO book?" And someone says, "Oh, buy it on Amazon at this link..."
Anyway, you guys know that I am still quite the abecedarian when it comes to this stuff, so this is more rambling than any real hypothesis.
And Google's smarter than you. they're also better looking. And have more money.
And google also spends their Saturday nights alone with their iPod, wishing they had a girlfriend...
Staying diversified insofar as the anchor text and rigid guidelines for where you choose to work with in terms of potential links covers 99.9% of the risk in my opinion.
Taking it any further is 95/5 and is something I do when I am at the top of the ranking heap and want to stay there.
I believe brand links are developed through signals.
One signal is to use your business / brand name often in your inbound links.
Using a link analysis excel which shows you all percentages for competitors along with exact keyword percentages and brand percentages helps. We tend to use an excel randomizer for anchor text selection except where we are not in any control of the anchor text we receive.
|Using a link analysis excel which shows you all percentages for competitors along with exact keyword percentages and brand percentages helps. |
That's exactly the kind of think that can get you into trouble long term.
And whether that happens isn't a matter of how smart one is. It's completely dependent on what Google does.
I know I am rambling again, but...
If I were a financial planner, I would tell my clients to diversify. Buy stocks, bonds, real estate, gold, diamonds, T Bills, pork bellies... just diversify for safety.
Which is fine.
But then the government decides that they are going to tax the "One Percenters" out there at a higher rate. They want to triple the tax rate - they just can't decide which particular individual financial instrument (stocks, bonds, pork bellies, etc.,) they are going to raise taxes on.
So instead of increasing the tax rate on any one particular instrument, they decide to take the easy way out and triple the tax rate on PEOPLE who make money from more than three different investment vehicles, regardless of what those instruments are.
When asked by reporters why they did it, a government spokesman replied, "Because it was easy, and we are lazy."
I can't help but wonder whether over diversification - in content, not in link profiles - is what caused so many sites to become Panda fodder...
|If I were a financial planner, I would tell my clients to diversify. Buy stocks, bonds, real estate, gold, diamonds, T Bills, pork bellies... just diversify for safety. |
and that's the wrong advice. you do diversify, but not randomly. There's statistics at work, you need a specific list of things that are correlated and not correlated and inversely correlated. When stocks go down, gold goes up, real estate is mostly unaffected (maybe not, but that's the idea).
This isn't investing. This is attempting to figure out Google's algo, while they're actively seeking footprints of people who are figuring out their algo.
So diversity is the key - no footprint. If there's one mantra, it's don't do anything twice. As soon as you do it three times, it's traceable by Google.
Which is why this:
|Using a link analysis excel which shows you all percentages for competitors along with exact keyword percentages and brand percentages helps. |
Is risky because as soon as Google changes their dials on attributes and what they're measuring, you're screwed. Diversify, and when they measure something else, well hey, I already have some of that.