Solving site architecture issues for SEO

by on 22nd February 2010

An excellent site architecture for search engines is almost exactly the same thing as great user experience: get to the content you need in as few clicks as possible. In this classic post, we explore the fundamentals of good site architecture.

Classic issues with PageRank distribution in site architecture

“PageRank” distribution is a classic issue that can impact your SEO efforts..

Imagine in the diagram below, the user (or search engine) enters your site at the homepage, and has to traverse a number of steps (or pages) to get to the content they’re looking for.

bad site architecture

The ranking threshold

Every step down the site architecture we take, we’re at risk of getting lost or losing interest in what we’re doing. This is the reason why I imagine Usability experts would normally always advise you to minimise steps along the conversion funnel to as few as possible.

Search engines don’t get lost but they are computing a level of authority that each page in the chain receives based on the link juice (or authority, PageRank, MozRank etc) that page is receiving from internal and external sources of links. In principle, the further down a navigational hierarchy a page is, the less authority it receives from the page above.

As Rand explains in this classic Whiteboard Friday on flat site architecture, “Google has a threshold” where, if pages aren’t receiving enough juice they may not be included in the main index of organic search results.

For a long time, this threshold was called the supplemental index, although I’ve not heard that term mentioned for quite a while. Regardless, there’s definitely a problem in the diagram above where, if there are too many steps between your homepage and the lowest tier of content, you could experience site indexation problems, particularly in very large industrial strength web sites.

A flattened site architecture

flattened site architecture

The solution to the problem: think carefully about the number of steps from top to bottom your navigation takes the user. This, very simple diagram shows a process of increasing the number of links from the homepage to the second level content tier, and the number of links from the second to the third, and so on.

At this point, I’ll remind you that I was talking about travel based sites, whose architecture very typically comprises of a continent > country > region > city based (vertical) hierarchy. With that in mind, though we’ve solved quite a major problem by flattening our site architecture, we could easily be introducing a new one.

A flattened site architecture with PageRank conservation in mind

flattened site architecture with pagerank

Let’s imagine for a moment that our homepage and some internal pages have attracted external links, perhaps enabling them to outrank their competitor’s pages. You may even observe these linked-to pages ranking generally higher than other, similar pages on the same domain. This is rather typical of any site and, in the diagram I’ve highlighted how the pages with more external links (and therefore more authority) might distribute their authority down the hierarchy to their lower tier child pages.

Why is this a problem? It isn’t, really. It’s pretty standard that some pages at the same level in your architecture will attract links and, when they do, they may be perceived to have more authority than pages at the same level. This is where I remind you that we’re looking at a specific vertical, travel – where a continent > country > region > city based (vertical) hierarchy can cause an issue that you may not have considered.

Vertical PageRank siloing

authority pages siloing PageRank

In this typical travel architecture example, we see our linked-to pages passing their authority down to their children. Obviously this is a very simple diagramatic approach, because some of that authority may be passed upwards to their parents through a breadcrumb. Just for the time being though, let’s assume that in our model, continents pass juice to countries in that continent, just as cities receive juice from their parent countries.

Solving the vertical siloing issue with a good cross linking strategy

Consider your cross link strategy to flatten site architecture.


Why not consider your horizontal, cross linking strategy as well as your vertical, site flattening strategy? By considering ways to link across vertical silos that otherwise may not receive much juice, you may be improving a significant chunk of your overall search engine visibility.

Ways to cross link your site

In the session I gave some general examples of ideas that would improve the cross linking architecture of your site. Those ideas included:

  • Similar destinations
  • Nearby areas
  • Most popular / top countries
  • Most popular / top cities
  • Recently reviewed locations / hotels / resorts

Of course there are many other ways to cross link the content on your site. Moving away from travel and into retail, popular products, similar products and “other users searched for” are all great cross linking examples. Then there’s UGC and reviews, employing user profile pages to indirectly cross link sections of content.

Ways to identify the most linked to pages on your site

Top Pages on Domain Tool - SEOmoz

If you’re looking for inspiration before you design a cross link strategy, try this methodology to determine your most linked to subfolders. Having an awareness of the most commonly linked to pages on your site architecture can really help fuel new ideas for cross linking.


  1. Great stuff Richard, and I’m glad you’ve written this up in such detail – it was a bit like condensing a whale into a matchbox during the session (if you’ll excuse the terrible analogy).

    I do have a question – what happens if you’re linking across branches via a big drop down nav menu or via selected footer links – does that limit some of the problems you describe above or does it just dilute the whole lot by having several hundred links per page? I would tend towards the latter but would be interested in hearing your thoughts. Definitely themed cross linking is the way forward in any case.

    • Hi Jaamit,

      In my experience the number of links you can use on a page really depends on what the links themselves are there to achieve. If you’re trying to boost authority at a mid sitemap level, I always say “don’t over cook your links” meaning try to start small and gradually increase your links until you’re happy with the metrics. 120 – 150 links per page? Totally depends on domain authority IMO.

      If we’re talking cross linking a lot of unique but very low level pages (jobs, routes, product pages etc) then sometimes you can have a few more links in there but you should definitely test this! I gave an example at the conference of using CSS based navigation techniques to “view all” links. Once you have a module like that on a page, you can gradually increase the links in the source code without damaging the UI. Keep testing until you’re feeling happy with overall site traffic / indexation metrics. On that note, I’d definitely work out an analytics strategy to keep a (daily) eye on changes like this.

      And remember – you’re better off starting small and increasing gradually rather than starting big and wishing you hadn’t :-)

  2. Hey Jammit,

    how would you link across branches via a big drop down nav menu and arent footer links given a lower priority over other links?

  3. Really good write up and really good to see this on the site as I didn’t manage to grab your session down at SES.

    Getting the internal link profile right this is something so often underutilised by many sites – and possibly one of the easiest to rectify. Its not a game winner by itself – but certainly its an important part of the team.

    • Thanks Pete. To be honest I think it’s easier to communicate a subject like this in a blog post, the time we’re given for this on a panel isn’t always enough to get the joyous complexity across in a simple manner: -)

  4. Got question ’bout crosslinking – which will be better – static links or dynamicaly generated links (or hybrid of that). If we choose static links then i think the “juice” will be better but if we choose dynamic then we could help google to index more “horizontally” subpages.

    In my opinion it was one of the most practical session at SES London. Thanks a lot for that.

  5. i love this post – particularly the diagrams.

    it would be great to see a follow up too on why you might *not* want to do this in some cases, or even why you might want to do the opposite.

  6. In my opinion this is essential when thinking about site structure nowadays. Great post, Richard!

  7. Great and timely pos, it would be very nice if you also could give us real world examples of sites with flat architeture and managing the cross linking in a good way, Amazon for example?

  8. In your diagram you don’t show cross linking at the very bottom level, was the intentional or was it just more work than was needed? Also are there any rules around cross linking between silos, should the links be relevant or does it not matter?

    • Hi Will, yep – got me there. I really couldn’t put any more time into the diagrams! As for the context in which it’s appropriate to cross link, I would try thinking about it from the perspective of a user. Does it make sense to see a link on a page? Try to find some relationship even if it’s ‘other users also visited…’

  9. Thanks for the response Richard. I understand from a user standpoint, but what about from a pure seo link juice passing standpoint? In your opinion does relevancy between linking pages based on each of those pages onsite (and offsite) factors matter, or is is just a numbers game?

    • I think when it comes to relevancy, the user’s in charge. There’ll always be an important argument for relevant internal (and external links) but at the end of the day there seems to be evidence (externally) that relevance has little impact especially compared to value driven metrics such as page authority.

      I think my best advice is, it depends. Do your own test and decide for yourself.

  10. Spot on, the diagrams are wonderful. Perfect timing with a project that I’m working on. The diagram showing the “hotspots” of page rank helps put it all into perspective, it’s not only the index page that drives juice around.

  11. Well, the way you had explained the flow of Page Rank in a website, it’s awesome. I have been many times confused about the concept of Supplementary index, or website is not being indexing properly, but now , the way you had explained page rank flow and site architecture, its very informative.

  12. I know Rich won’t mind me mentioning another post on the subject so for those who are interested in more commentary / diagrams, Rand has a great post over on SEOmoz Re: Indexation and Authority Distribution (link below)

  13. Richard, in your experience with large scale architectures have you ever seen Google devaluing links in sections, i.e. applying block analysis to the page layout and identifying linking sections that might be sitewide navigation? There was some talk a little while back about links in footer sections, for instance. This would have a definite impact on how you cross link your deeper pages to get maximum benefit. Many of the examples that we see of cross linking tend to put the links in easily identifiable blocks within the page. Just wondering if you’ve seen the big G treating these differently?

    • “linking sections that might be sitewide navigation”

      Yes – definitely. There’s no doubt in my mind that boiler plate code sections such as footer links and sidebar navigational elements pass far less value owing to the overall across-site duplication. Exactly how that works is up for debate – for example the first code block (say, on the homepage) could pass value and all the others coould be filtered, or it may be that every other link passes a diminishing level of value in some way.

      Certainly domain diversity plays a big role in this and that metric alone may be what drives my feeling that hugely duplicated links pass little to no additional value.

    • Well, if we look at it from an engine perspective and treat the links as votes, even internal ones, then sitewide nav actually shouldn’t be counted as votes for those pages perhaps, but in-body links perhaps should be since they are specific references. However I can also see the point that navigation typically only references your most important pages. I’m thinking that you might be right about the possibility of identifying nav areas and assigning value to the first one only.

      I wonder if anyone will be brave enough to nofollow duplicate links in sub/alt nav areas to test this out!

  14. Great post, I like the diagramms…

  15. i like the diagrammme very much keep it up

  16. Wondeful information about site architecture. The diagrams were a big help in understanding the different ways of link building, whats benificial and what can hurt you. Thank you.

  17. Thanks for this article on the ideal ‘pyramid’ seo sitestructure. Good stuff!

    In my mind I often come across a ‘multiple pyramid’ structure though, like:
    1) Products: red blue white widgets
    2) Articles: red blue white widgets
    3) Blog: red blue white widgets

    What is your take on this? Is this really different?
    Would you recommend rearranging it into one big pyramid, e.g. placing the (how to) articles under the related products?

Comments are closed.

We're hiring – check out our careers page Careers

Get insights straight to your inbox

Stay one step ahead of the competition with our monthly Inner Circle email full of resources, industry developments and opinions from around the web.