The best internal linking structure depends on your business model
Opposite to common belief, internal linking best practices are not commonly applicable but vary by site type and business model. This article shows what model to apply in which case.
A building is not a building. A skyscraper is built after a different philosophy compared to an airport. Internal linking is no different. The right structure depends on the type of site, i.e. business model. Not every site is alike, but we often apply the same tactics to internal link optimization.
And that’s the problem.
The centralized vs. decentralized model I explain in this article forms the basis for choosing the right internal link optimization strategy. This model is a guideline for distributing PageRank and CheiRank throughout a site and a prequel to TIPR.
A primer on PageRank and CheiRank
Let’s think about the goal of optimizing internal linking for a minute. Besides making it easier for users to find what they’re looking for, internal links make it easier for search engines to crawl every page on a site, so they can index the content and show it in the search results. But for search engines, it’s not just about finding all pages but about additional ranking signals, a.k.a. internal PageRank.
As you know, it has become harder to build backlinks nowadays. Thus, we need to make the ones we get count. The balance between PageRank (PR) and CheiRank (CR) is a big differentiator between using a site’s backlinks efficiently and wasting them.
As explained in TIPR, PR is the authority a URL receives, CR the authority it gives away. Both can be measured in a value between 0 and 1. PageRank is passed from page to page and site to site.
We need to be wary that some pages of a site end up with more PageRank or CheiRank than others. This creates two imbalances:
some pages have a higher or lower PR and CR than others
the relationship between PR and CR of the same page can be skewed
Note that the concept of some pages to “hoard” PageRank seems to be leaky and possibly outright wrong. The goal of this model is to distribute PageRank strategically. CheiRank helps to measure whether pages with high PR give away enough to other pages. PageRank helps you measure whether important pages have enough PageRank.
Even though the impact of backlinks has become thinner over the years, PR and CR still play an important role in ranking pages and sites. Thus, we should optimize how PR and CR flow through our site.
To answer that question from a 1000 feet perspective, we must look at the type of a site.
There are two types of websites
Websites vary by the point of conversion, depending on what business model they follow. One type of site leads all users to one or a few landing pages, the other has users sign up on almost every page. Should the approach to internal link optimization be the same for both? Of course, not!
Sites that have a few points of conversion are what I call “centralized”. They don’t have scalable page templates, in most cases. Instead, they consist of landing pages, a blog, and some other pages. Centralized sites are often used by SaaS, app, and enterprise companies like Atlassian, UBER, or Salesforce.
The opposite is sites with many points of conversion, used by Ecommerce businesses, social networks, and marketplaces. They are “decentralized” sites with page templates they can scale, such as public instances, user profiles, apartment listings, products, or categories. Examples are Pinterest, Airbnb, and Amazon.
Centralized Internal Linking
The goal of centralized sites is to focus PageRank and CheiRank on landing and a few assisting pages. It’s not that onlylanding pages should receive PageRank, but mostof it. In some instances, you want a blog article or resource page to be strong as well because it targets a competitive keyword.
There are different kinds of landing pages:
Product landing pages
Often, product landing pages target the product name (“Jira”), a brand combination keyword (“Jira project management”), and a few generic keywords (“project management tool”, “bug tracker”). It’s not too difficult to rank for the first two, but in order to rank for short head generic terms, PageRank often is a differentiator. Now, people hardly link to product landing pages, so in most cases, you need to funnel PR internally to the product landing page. Thus, PR and CR flow are not balanced, they’re focused. Not spread, but centralized.
Decentralized Internal Linking
The goal of decentralized sites is to distribute PageRank and CheiRank equally since conversions happen on almost all pages. Pinterest has millions of boards, Airbnb has millions of apartments, Amazon has millions of products. Unregistered users can sign-up on any page they find on Google or through an ad. The site structure doesn’t follow a funnel in the classic sense. It’s best to strive for a balance of PageRank and CheiRank.
Make no mistake, some marketplaces have landing pages to address one side of the market. An example is Upwork, which has search pages for freelancer professions or categories for searchers and sign-up landing pages for freelancers to provide their services on the platform. However, in this case, the landing pages get a lot of PageRank from other pages.
Decentralized sites also benefit largely from having a lot of pages, of which most are usually relevant. That increases the chance of getting backlinks and providing for signals for relevance. 10 internal links pointing at a page and saying it’s about a specific topic is better than 3, as for centralized sites.
News sites and publishers fare best with the decentralized model because readers can sign up for a subscription or see an ad on every article.
Optimizing with the model
When optimizing through the lens of the centralized or decentralized model, there are a few points to consider.
First, it’s important to understand that you’re never going to get a perfectly focused or balanced distribution of PR and CR. It’s always just an optimum to strive for.
Second, to monitor progress and prioritize URLs correctly, apply the TIPR model, which combines PR, CR, backlinks, and log files. After my mind, log files are the best tool for that.
When crawling your site, as described in the TIPR process, group the pages by template or type. Then look at which group is on average over- or underfunded in terms of PageRank and CheiRank.
The tool you use doesn’t matter much. Here are a few options:
Third, your goal is not to find a few links you can add manually but find opportunities to change internal linking at scale. That means you’ll often adjust the top-nav, side-nav, or footer elements, also known as programmatic internal links.
Fourth, you should analyze your internal linking structure regularly and make adjustments as your site adds new pages and backlinks over time. Look at your log files on a weekly cadence if possible and reassess/optimize internal linking on a quarterly basis. Leave some room between rolling out changes to internal linking so you can assess its impact.
A site falls in either category, centralized or decentralized. Being aware of the difference should guide you when optimizing internal linking.
To visualize the two models, I crawled www.amazon.com and compared it with a crawl of www.atlassian.com. To keep it at a reasonable size, I crawled 10,000 pages of www.amazon.com and the full 4,400 pages of www.atlassian.com. I used Screaming Frog with “user-agent=Google Bot Desktop” and JS wasn’t crawled. The metric visualized is “link depth” (number of clicks it takes from the homepage to the respective page).
Amazon link depth
Atlassian link depth
As you can see, Amazon’s internal link structure mimics many small islands while Atlassian’s is much more “inside out”. Amazon is a decentralized site and spreads links over many categories. Atlassian is a centralized site that keeps PageRank “close to the heart”, meaning focuses on landing pages and close other ones.
If you haven’t done so, read my article about the TIPR model. Please share your experience from applying both models with me, as I’m continually trying to improve them.