Everything you need to know about SEO, delivered every Thursday.
The Problem With Identifying Problem Links
Since the last Penguin update, there has been a lot of chatter about examining your link profile in order to identify bad links. Whether you’ve been hit by that update or your site remains unscathed (for now), the potential danger of “unnatural” links is top of mind. Many people who’ve never given a second thought to their link profiles are running backlink reports and trying to identify whether they stand a chance of being hit by a future update or manual penalty.
However, if you’ve ever used a tool like Link Detox or LinkRisk (for the record, I do use both), you might have noticed something: some of the links identified as being the worst offenders are actually good or even great links. Sometimes, the links that you’d pick out as being the bad ones don’t get flagged at all.
It’s the same problem that any algorithm faces: it’s impossible to determine whether a link is 100% good or bad, because the metrics don’t tell the full story. These tools definitely help us narrow down the list of potential offenders, and when you’re dealing with a profile that contains tens of thousands of linking domains, it would be exceptionally difficult to wade through those manually without any help.
However, once your tool of choice has identified potential suspicious links, it is important to review this list with a critical eye. For each link in question, ask yourself: “Is this link good for my site?”
The Problem With Relying Solely On Tools
When I was beta testing LinkRisk, I ran one of my sites — a music site — through its system to see if we had any bad inbound links. There have been no intentional link building efforts for the site and no paid links — generally not much more than content creation and social promotion going on. In my naive mind, I assumed this meant that we shouldn’t and couldn’t have had any really bad links.
The results were surprising, though. A link identified as being extremely toxic was one that I would consider among our best — it brings us relevant traffic, and it was editorially given from a thematically related site (in this case, a local music site).
A second link identified as being a very poor one was from the local college radio station where my site’s owners host a weekly radio show — another relevant link that sends us nice traffic.
A third is from a music venue where we have sponsored shows and do a lot of offline marketing.
Actually, almost all of the links identified as the worst ones are totally natural and good for the site. They’re from websites of bands we’ve interviewed or local partners we’ve worked with. The site has never been penalized or hit by any big update, but if I relied solely on what this tool told me, I might pursue removal of a few of the site’s most valuable backlinks.
Another issue is that the two main tools that I use for link risk management don’t always agree on which links are suspicious/unnatural:
My site’s link profile is high risk according to LinkRisk and low risk according to Link Detox. Different data source and proprietary scoring algorithms are obviously among the reasons for the discrepancies, but this still leaves me confused. How is your average webmaster supposed to make an accurate decision about which links might need to be removed or disavowed?
What Should You Do?
Figure out what it is you want from a link. For one of my sites, I want local relevance and a sign that referral traffic has an interest in our content. For another one, I want someone to fill out a contact form (and hopefully sign up for something). For another, I want good search engine rankings. Knowing what you want to accomplish helps you determine what to do.
Visit the sites and look at the links. There’s obviously a reason why they were identified as being risky, so see if you can tell what that reason is. For example, one of my main “bad” links is from a site-wide blogroll on a massive site. It may seem “spammy” at first glance, but every blogroll link on that site is relevant to the site itself. Is that why this inbound link is being flagged? I’m truly not sure, but what I do know is that there are many good links getting flagged as suspicious. (Link Detox does provide a reason for each flagged link, however, which is quite helpful.)
Find out if these links are sending you any traffic. Look at the other stats for that traffic, too, especially the conversion information (if you track that) and metrics like pages viewed per visit or time spent on the site.
One of my “bad” links sent me visitors who clicked through several pages of the site (5.67 on average) and that’s an important metric to me for this site. We’re local, and we want people to click through a few posts and read what we have to offer. I’d rather see 70 people viewing close to 6 pages each than 1000 people landing on the site and immediately leaving.
The Bottom Line: Use Your Brain
Relying on tools alone is a poor way to analyze links. Different tools use different data –and if you’ve ever run a report in more than one tool at a time, I’m sure you’ve noticed that the data rarely match up exactly. Conversion data on AdWords doesn’t always match what is pulled in through Google Analytics.
Heck, I can’t even get the Majestic API data to match what I get when I run a site in their own tool. If you can’t get the same result twice, why would you trust it completely to make decisions that could impact your site in the future?
I’m not at all dissing these tools, either. I think they’re invaluable, but I also think they need to be used with caution. If you have been penalized or badly hit by an algorithm update, I’d highly suggest using these tools to see where your problems might be, because from my experience, many webmasters and site owners simply are not able to identify all of their problem links for various reasons. I have yet to analyze a link profile that did not contain loads of bad links that were still live after a webmaster or business owner has gone through and done cleanup.
So, use these tools… but use them wisely.
Some opinions expressed in this article may be those of a guest author and not necessarily Search Engine Land. Staff authors are listed here.