Silo Web Design

Silo structured web sites are an idea that was popularized by an ebook from Charles Helfin.

SEO Basics

Search engine optimization mainly comes down to two things: The document itself, and its whereabouts in a network of hyperlinked documents.

On the one hand, search engines use algorithms related to semantics. They’re able to extract key ideas from a document taken out of context, based on how information is organized within it.

On the other, how documents are linked to one another lets search engines interpret a context, and come up with SEO factors such as a site’s authority on a topic.

There are other factors, of course. The web evolves and these are introduced in a feedback loop as the data gets reprocessed. These include:

  • Time (e.g. how long a document has been around)
  • End user interaction (e.g. do users actually click it when it gets listed in search engines)
  • Technical factors (e.g. IP address) that search engines use to filter out spammers

Ways to bend how a search engine puts documents in context include:

  1. Campaigning for relevant incoming links to your site
  2. Increasing the number of documents on your site, and pushing a few of them forward by using a well-rounded internal linking strategy
  3. Focusing outgoing links from a particular document onto other documents on the same topic

In practice, campaigning for links means buying text links and picking up your phone to convince webmasters to link to your site.

Pushing documents forward is reasonably straightforward to automate, and this is what Semiologic Pro has been doing from day one.

Focusing outgoing links, which is the main idea behind silo structured websites, is much more easily said than done. It is by no means easy to automate. And in my opinion, two factors should rule out excessive outgoing link trimming (and the use of Posts and WordPress categories).

Firstly, a site needs to remain as easy to navigate as possible. This goes down to basic usability, and this ultimately translates to SEO in that users will stay longer on the site.

Secondly, the better a site is indexed overall, the better it fares for the documents you’re pushing forward. This is basic SEO, and an overwhelming argument to keep some links around to boost the ranking of less important documents.

Optimizing The Site In Practice

The Silo Widgets plugin introduces three widgets:

  1. A Silo Pages widget. Use it in a sidebar to output what amounts to an automatically generated navigation menu for your static pages.
  2. A Silo Stub widget. Use it as an inline widget when you’d like to automatically generate the contents of a page whose only purpose is to link to its children.
  3. A Silo Map widget. Use it as an inline widget on a site map page, to automatically output a full site map of your site.

To use them, browse Design / Widgets. Add a Silo Pages widget to one of your sidebars (or a Nav Menu Widget, if you need more flexibility), and the other two in the Inline Widgets “sidebar” (see the inline widgets feature) so they become available in the page editor.

Following on, build your site using the following strategy:

1. Put Your Main Content On Static Pages

Static pages (Write / Page) let you create content that exists outside of the blog chronology.

Contrary to posts, static pages can be organized in a hierarchy using page parents (to the right in the editor). Your site structure could look like this:

/
/section1
/section1/topic1
/section1/topic2
/section2
/section2/topic1
/section2/topic2

… which, by the way, is the most natural way to organize a site.

As seen on this site, the Silo Pages widget (“Browse”) will display:

  • The page’s children and siblings
  • The page’s parents and their siblings

Otherwise put, the Silo Pages widget doesn’t do things “by the book”. It also adds the element of usability that will keep the site fresh and navigable.

One argument against static pages is they don’t have categories. Ignore it. In practice, you don’t need them, and the Silo Widgets plugin lets you add page tags. Related Widgets can then build on the latter to generate lists of related content.

Two bonuses related to using static pages for your main content include:

  1. Your site is much more navigable — not to mention maintainable
  2. Your content becomes splog-proof, since static pages don’t show in RSS feeds

This brings me to…

2. Use Blog Posts To Link To Your Static Pages

SEO-wise, the power of blogs comes from the constant stream of updates. To take full advantage of it, use your blog to:

  • Maintain some kind of news section related to your topic of interest
  • Announce new pages and updates to existing pages

Mention your site’s content (static pages) in your posts, and link to it. Use a few tags for each post.

Be selective however. Not everything belongs on a static page…

  • If you’re writing a project management tutorial, a static page is appropriate; announce it in a post, and link to it whenever it gets updated and when you write about the stuff
  • If you’re merely reacting to a news article on risk analysis and giving your opinion, it’s time sensitive, and a simple post is best (with a link to that static page you wrote on the same topic, of course)
  • If you’re running a podcast show, use posts. It’s all news and you want your media to show up in your RSS feed

A Word On LSI

Contrary to a belief that recently became wide-spread, latent semantic indexing (LSI) is NOT used by search engines nowadays.

The idea has been around in search engine research for years, and it is very theoretical. In layman’s terms, here’s how it works:

  1. You scan a page for keywords and key phrases
  2. As you do this, you mark key phrases that seem to “work together” (e.g. dog training is related to dogs and dog trainers)
  3. You do this over and over again for a multitude of pages
  4. You combine a result to come up with a matrix that links word or expression A with word or expression B based on whether or not they’re usually found together

This sounds great in theory for the unexperienced eye, but it’s not that great in practice. It requires a obscene amount of computing power to do it properly.

Keep in mind that most languages have over 200k words. And while basic English is relatively easy to learn, proper English is harder to master than Chinese. English is the language that features the most unique words of all (twice as many as French).

To do proper LSI, you thus need to process a matrix with one line and column per word, i.e. 500k squared pieces of information… Google’s Grid (well over 10k computers working together) is nowhere near able to manage this amount of information in a reasonable time.

Pollute the meaningful results with all sorts of stop words (words like “the” or “a”) and expressions (e.g. “to turn up” has a very different meaning from “to turn down”), and hell breaks loose.

The truth is, we’re nowhere near able to use LSI in practice.

Here’s what happens in reality: Google and other search engines identify and maintain some level of LSI by investigating the keywords customers purchase ads for. For instance, Dog training sites will usually buy keywords for dogs and dog training.