Two weeks ago, I had a chance to present on Actionable Metrics and Diagnosis at SMX West along with Vanessa Fox, Conrad Saam, and Jim Yu. It was my first time speaking at SMX West and was a great experience, so a huge thanks to SMX for having me speak.
I want to share my slides and add notes for everyone who couldn’t make it or wanted more details.
There are two great sources for link data, which are Majestic SEO and Open Site Explorer. We’ve also built out a number of internal toolkits using the SEOmoz API, which help us scale data pulls and analysis.
I showed an example of our hello(KITTY) toolkit. HELLO stands for Here is Every Link Linkscape Offers. This is one part of our toolkit that allows us to drop in a list of URLs and export all of the SEOmoz metrics into a CSV.
There are a few quick and easy examples of how this can be used:
- Guest Blog Posts: After mining a list of domains for guest blogging, you can drop in the domains and sort and bucket by Domain Authority. This bucketing can be used when doing outreach.
- Directory Submissions: Identify the URL you’ll receive a link from, process them to check for indexation / cache. Drop cleaned list into tool, sort by mR or PA to prioritize submissions.
- Competitive Analysis: Drop in a list of competitive sites to perform competitive analysis on the SERPs for targeted terms.
Averages and Spread
I spoke briefly about how to perform competitive link analysis. I moved through this very quickly in my presentation, because I’ve written on competitive analysis elsewhere. I discussed how you want to remove large domains and exact match domains, as they may skew analysis. You want to calculate averages and standard deviations for all of the major link metrics.
The goal here is to take your site, line it up against these averages and evaluate these metrics to see where you’re over- and under-performing. I’m pretty sure I used the phrase “over-optimized” around this point and was called out on Twitter for it. I stand behind the usage of that word, even if it’s semantically inaccurate.
Important Competitive Metrics
I went a bit deeper with examples of a few key competitive metrics by lining up an example site (in dark blue on slide 6).
Actionable takeaways from this example analysis:
- The site is falling short on average number of linking root domains. The site needs more links from more domains and this gives you a vague gauge of how much.
- The site is doing “ok” in its targeting, since it had 30 LRD linking with the exact anchor, but it’s falling short relative to the average.
- The site needs to be cautious when getting more anchored links, because their anchor text distribution shows they’re high relative to the average. I’d focus on increasing links with secondary terms, variations, and branded terms. This will allow a site to drive up LRD without being overly aggressive on that particular term.
The standard deviation helps describe the spread of the data across the top 10 (is it tightly packed around the average or is it a wide spread). This will let me know when I need to investigate further.
To investigate the spread further, I can visualize it by graphing it out inside of Excel. This will show me a few new insights.
- The majority of the high average comes from the top two sites. The sites from 3 to 10 are not very competitive (opportunity exists).
- Top sites are aggressively targeting this term. I wrote more about why this is a bad thing in my competitive analysis post.
So far, I focused on numbers, but often you want to perform some type of qualitative analysis on links. There are a lot of reasons you want this information, but I gave a few examples.
- What type of sites / pages are linking?
- What’s the quality of that link?
- What’s the relevance of this link?
You often want to pull a link report and manually dig through a list of links in Excel to check. This is hard to do at scale, so what’s a good solution? Build a scraper.
Build a Scraper
This script will take an Open Site Explorer CSV, run through all the URLs, download the content, parse it for information, and append that data into the CSV. The output is a CSV of OSE link data with additional columns for qualitative data.
Profiling Link Type
From there, I made a pivot table in Excel. In the example given, I checked for the CMS running the sites linking to Distilled.co.uk. This quickly shows we have a large number of WordPress sites (blogs) linking, but not a large number of forum CMS. Maybe Distilled should put more effort in participating in forum communities.
This is a scalable way to check for qualitative metrics.
Link Velocity Trends
I like to monitor a site’s link velocity trends. It gives some insights about how links are being built. It gave me a big ah-ha! when I was researching the Mormon LDS Church link profile. (I had a chance to meet the whole LDS SEO team at SMX. They’re great guys with some smart strategies.)
The best source for this data right now is Majestic SEO.
- Natural patterns: Is there a large spike upfront, then dies off? Is there no growth for a long time, then moves up to a steady rate? Are there spikes? What’s the reason for spikes and growth?
- Temporal Analysis: Be able to dig into a profile by date and see what links were discovered to cause increases and spikes.
Protip: I made a comment here that was touched on in the Best of SMX post from Gil Reich. I talked about how your link over time graph should make sense. That link increases should have a corresponding reason and that I’ve seen value in the synergy of running content / branded link building at same time as more aggressive link building tactics. Although I wasn’t referring to paid links with that comment.
Shortcomings: These are just volume numbers and discovery dates. These numbers are easy to manipulate with naive link building methods. A large spike may not be indicative of a serious gain.
Real Link Growth Rate
By pulling a link report from Majestic SEO, you can create a pivot table to sort links by date discovered. These URLs can be matched up with their ACRank and summed by date. This can be replotted over time. Spikes now show points where volumes of ACRank were gained, not just links in large numbers.
You can expand out dates in the pivot table to find URLs discovered on specific dates.
Paid Link Penalty
I decided to wrap up with a real life example of solving a penalty associated with paid links. I used JCPenny as an example.
Disclaimer: JCPenny is not a client. They’re just an example and you can get this data at Open Site Explorer.
I pulled a link report from OSE and made a pivot table that shows URLs linking by Anchor Text. For each URL, I showed the PA of that URL and summed up the PA per anchor. This shows the summation of PA of all URLs linking with a specific exact match term (you could think of this as a vague metric for the amount of value passed per term, but I know this isn’t 100% accurate because a lot of factors can influence value passed). I did this to find interesting outliers in the link profile.
What I found were terms like “Twin Bedding” scoring as high as terms like “JCPenny”,”here”, and “www.jcp.com”. When you expand out the pivot table for that anchor, you get a list of all the URLs linking with that anchor text. Upon inspection, all of them are linking in the fashion shown on the slide. This looks a lot like a paid link.
So this would be a quick way to identify odd anchor text outliers and walk away with an action plan of domains to approach to have links removed or cleaned up.
I wrapped up by announcing our Excel for SEO Guide. If you haven’t seen it yet, you should. It covers a lot of great information on using Excel to dig through piles of data.
If you have any questions about my presentation, feel free to ask in the comments. I enjoyed speaking and look forward to hopefully speaking again soon. And if you’re in Seattle, I’ll be speaking on WordPress SEO at Wordcamp Seattle on April 16th. As always, you can find me on Twitter if you’d like to chat more.