Stuck In Link Limbo? How To Craft Real Business Metrics For Link Building

As the in-house person in charge of marketing the online legal directory, Avvo, I need to have accurate and timely access to all of our performance metrics. Since we launched the site three years ago, finding consistent metrics to evaluate our site’s growing link strength has been a remarkably difficult exercise. We have adopted at least three different metrics and used three different tools, yet I remain frustrated by our inability to develop a consistent model to evaluate our site’s link performance.

From a business perspective, something that feels so easily quantifiable should fall into the simple process of setting objectives, defining benchmarks, and measuring results.  But it’s not that simplistic. SEOs get tied in knots speculating about what impacts the quality of a site’s link profile. The potential variables are endless — domain diversity, bounce rates, anchor text, absolute number of links, relevancy, redirects, nofollows (or not?), internal links, paid sponsored links, deep links, non-profit links, density of links on linking page, education links, international links, links written by left handed atheists, social media links, age of links, links from Malia Obama’s Facebook page,  and so forth.

Finding simple business metrics to accommodate all of these variables (and even knowing if they are the right ones) is extremely difficult. For a comprehensive overview of tools available to help keep track, check out A Big Roundup of Linkbuilding Tools by Debra Mastaler. The group I’ve gotten my hands dirty with include:

• Site Explorer – the granddaddy of link checking tools, provided for free by Yahoo. Admittedly, this was our starting point at Avvo, but the data here was so erratic and incomplete, it was functionally impossible to use. Additionally, Site Explorer has been slated for the digital scrap heap.
• Google Back Links Checker – Big G offers such a small sample of actual links that doing any competitive benchmarking is impossible. Even Google webmaster tools doesn’t show all links to your own site. While it recently updated to open up even more of the link profile kimono, you still don’t get 100% accuracy; although it shows much more comprehensive data than the link: operator.
• Digital Point – offers a backlink tracking utility, which it accesses through Google’s API, so it has the same problems as the Google link: operator. Unfortunately, it only shows the raw number of links and only tracks individual page URLs, nothing in aggregate.
• Linkscape – pay-for-tool from SEOmoz goes further than anything I’ve seen available to evaluate not only the quantity, but also the quality of links. Linkscape data is pulled from SEOmoz’ own index of the web. SEOmoz recently launched a new free product on top of Linkscape called Open Site Explorer.  Matt McGee covered OSE launch for SEL.
• Majestic SEO – fairly comprehensive link analysis based on review of a proprietary URL index updated monthly. They offer a free version, but you have to verify ownership of your site (no thanks, I already have enough third party stuff on our site.) One thing Majestic does provide (for free) is a handy dandy graph of your link history.  Here’s the graph for Avvo:

Both Linkscape and Majestic up the game on other tools by emulating the search engine crawlers and developing their own indices of links across the web. Majestic provides a monthly re-index. SEOmoz’s updates have been a little more erratic, but I believe they have settled into a more regular schedule recently. The problem with these third-party indices is that they sample different portions of the web every time they put out a new index. This means the absolute number of reported links may go down as they index fewer pages across the web (to deal with duplicate content or spam for example) or up – as the crawlers find more ancient pages.

The data fluctuations due to constantly changing indices wreck havoc on simple business processes. Note in the graphic above Avvo’s rollercoaster – with fluctuations of at least 10% every month in the second half of ‘09. Based on the reports here, we’ve vacillated between wild improvements and shocking losses every single month in the last two quarters of ’09.

If I had set an objective of improving links, defined a benchmark and then measured against that benchmark using these tools, I’d be pulling my hair out (see bio image.) If I was paying a vendor, I’d be screaming about my retainer, or thinking of extending it further. If I had a linkbuilding employee, I either would have fired her or over-bonused her. If I had to report back to my board, I’d be either boastful or sheepish depending on the month. You get the point: in every case, the erratic reporting leaves me entirely unsure.

When counting links isn’t enough

Simply counting the number of links and domains reported by these third party tools clearly doesn’t work. To address the erratic data issue, we shifted our link tracking monitoring metrics to a more qualitative approach provided by the SEOmoz tool. Linkscape takes link analysis a step further and provides their own statistical evaluation of the quality of individual links and the overall domain’s authority, along with a slew of other quality-type metrics. The primary metric is called mozRank, which is a 1,000 point scale and essentially is built to reflect SEOmoz’s interpretation of PageRank. This, in theory, is the ultimate solution for the business process of tracking authority growth. I can compare our progress with our competitors at very granular increments

Unfortunately, theory and practice collide.

First, there are so many different metrics to choose from within Linkscape, I still can’t discern the differences between all of them – we settled on Root Domain Domain mozRank. Second, we’ve experienced metric fluctuations that just don’t make sense. As a rapidly growing startup, our domain authority is constantly increasing, yet we’ve seen down cycles with mozRank, including one incident when our mozRank dropped .04 to 5.83 on the same month our visible Google PageRank moved up to a 7. Granted, this is a very small change, but when you offer a very precise instrument, you expect very precise results, not to mention the numbers are a full logarithmic step apart.

Additionally, the search algorithms change. Assuming the statistical minds at SEOmoz can reverse engineer authority profiles, the cycle of collecting data, validating data and then tweaking their own algorithm means they will always be way behind the here and now. Finally, the bigger issue here, of course, is that PageRank is deliberately fuzzy, so at best, SEOmoz is trying to match a deliberately inaccurate number.

I’ve also tried benchmarking against competitors by creating performance ratios. Using Linkscape data, we tracked our absolute number of links and number of linking domains against an established competitor. This works on the theory that while the data from Linkscape and Majestic may fluctuate based on their most recent indices, the relative reporting for sites within a business category should be similar.  The data is similarly bad across sites and therefore you can track overall performance improvements. Additionally, a competitive ratio as a KPI tells you how far behind (or ahead) you are.

Again, practice collided with theory. Our benchmarking ratio graph looks like this:

The matrix below shows wild fluctuations in Avvo’s ratios month to month, including one period when Linkscape reported a loss of over 500,000 links to our competitive benchmark moving the ratio from 4.5% to 10.9%. This is just far too variable to use as a month to month managerial tool.

What does count when it comes to building links?

So I found myself back at square one. The myriad variables associated with links makes finding a solution imprecise at best. However, I’d offer an obvious proxy solution for measuring link performance: referring traffic. Link building in its purest sense is a business development activity.

Thus, growth in a site’s overall link profile should be reflected in strong growth in referring traffic. It’s not a massive leap of imagination to presume that links that deliver traffic are more algorithmically valuable. And if you are playing by the rules, you can filter those paid (and appropriately nofollowed) links in your analytics package. Below is a snapshot of a link to our “How to Date a Lawyer” page that generated great traffic (and therefore SEO value) to Avvo.

Presumably, this is the type of link that really impacts site authority and the success you want to see from your link builders. Referring traffic is neither perfect nor particularly granular. It does, however, provide a workable framework for running the business process of link building by either employees or consultants.