Skip to main content
  • Agency for web development
  • Content Management with Primer
  • Open Source Leadership with Drupal
    Language
  • Deutsch
  • English
  • Contact
Site logo
Agentur für Webentwicklung
  • References
  • Services
  • News
  • About us
  • Agency for web development
  • Content Management with Primer
  • Open Source Leadership with Drupal
close

How we use graphs to analyze website structure

12. February 2021
Mathilde Dumond Porträt
Mathilde Dumond

There are many articles out there to help you reflect on the structure of your website, in order to improve User Experience or SEO ranking. Most of them focus on rules to follow, and rather empirical decision-making processes. They are very useful, but lack a dimension about analysing the current situation of your website, and improving it.

We thus started to develop a solution allowing us to visualize the current structure of the website, this way creating a data-driven process to improve both User Experience and SEO rankings.

User experience of a website and SEO depend on many factors, such as design and page organization, but also the architecture of the website, i.e. how are the pages linked together.

This can already be assessed through the structure of the menus, but a lot of navigation can also be done through content links. There is currently no satisfying solution that allows a complete and intuitive visualization of this structure in relation with relevancy data.

Display the structure of your website in relation to relevancy data with our solution

After a first successful prototype we decided to develop a solution allowing our clients to assess the structure of their website. Because the table based approach of relational databases is not  helpful to understand the bigger picture, we first map the website onto a graph database in Neo4j. This is useful because a graph allows a more intuitive visualization of the structure of the website, and of the relationships between the pages. One key advantage of using a graph Database will be the computation of the depth of pages, i.e. the minimal number of required clicks to reach a page.

Then, because Neo4j does not have suitable visualisation capabilities, we use the javascript library vis.js to visualize the networks. It simply allows us to customize the representation of the graph, to manipulate the physics of the model, and overall to make the visualization easier to understand while carrying more information.

This framework allowed us to give insights to our customers on how to improve their website’s structure, using a data-centric approach.

Our website as an example

To protect the privacy of our customers, we will use our website as an example to showcase this tool. The method provides orientation in our small website and is even more helpful in an environment with higher architectural complexity.

We use shapes and colors to carry information on the graph, and allow an easier orientation. For example, we can associate a type of page (belongs to the Blog, is a News article, a Profile, etc) to a color and thus recognize the semantic structures of the website. This visible structure can sometimes be associated with the various personas of the website.

In our case, the website has two main regions: the documentation, separated from the main website, which is mostly composed of our case studies, our blog and the features of our platform. We can then clearly separate the “client” from the “prospective client” personas.

structure of our website
Structure of our website. Using colors makes it easy for the reader to recognize major structures on the website. In this case we can see a parallel organization between the two main areas of the website: the Feature pages and the Documentation pages, which is expected since each Feature is linked to the page corresponding to its documentation.

It did not happen on our website because it is quite small, but on one of our customer’s bigger websites it happened that clusters of pages were not reachable at all on the website. Because the pages were still visited (probably directly from external links), it would have been impossible for our client to find out this problem without using our graph-based analysis.

Visualizing relevance data

We can adapt the size of the nodes to the number of visits at any moment in time. This allows us to detect clusters or groups of over- or under-visited nodes on the website. While Google Analytics or Matomo analytics tools do allow you to access the most visited pages, this representation has the advantage of displaying groups of related pages that have higher or lower visits. This can give more information about your users centers of interests, and help you to create content that your users will find interesting.

Visualizing page depth in relation to relevance data

Graphs are also very powerful tools to compute distance between pages, task where relational databases strongly underperform. With graphs, this can be done very efficiently, and this allows us to determine the click-depth for all pages in the website. For example in our website, this allowed us to detect that the former front page was still published - but unreachable.

click distance and relency data on our website
Pages depth relationship to visit numbers on our website. Color of the dots is the minimal required number of clicks to reach the page from the front page. The black dots are unreachable and include the former frontpage of the website, and a few pages linked only from it. This study will allow us to improve our website by deleting the unreachable pages and redirecting users, and make sure that all important content can be reached.

Coordinating the depth metric to the amount of visit for any page can help you detect pages with few visits that are deep in the website, making them difficult for the users to find.

In addition, it is also possible to compare visits numbers, for example between smartphone and desktop visits, compare time periods, or even find out which pages are visited through google searches compared to direct traffic.

What about me?

This solution is so far tailored to websites developed on our platform Primer, but we are working on coupling it with a crawler to enable the analysis of any website. This of course would however not allow the detection of unreachable pages, unless the interested party provides us with a list of urls to be tested.

We presented our solution to the DACH meetup last week and are grateful for the great feedback, read the presentation with more representations of our website below (in German).

If you are interested in us analysing your website, do not hesitate to contact us!

Digital sovereignty and open source

Miro am DigitUp 2025
19. May 2025

MD Systems stands for technology that serves people and has been doing so for 25 years. We are committed to ensuring that the digital environment is understandable, accessible and trustworthy - for everyone.

Mehr lesen

Drupal Mountain Camp 2025 in Davos

Kongresszentrum Davos Front
24. March 2025

This year's Drupal Mountain Camp was a complete success. Not only because we were able to take home another Splash Award, but also because of the people from the Drupal universe.

Read more

UNICEF Switzerland and Liechtenstein: Sustainable IT performance with MD Systems

27. February 2025

We supported UNICEF Switzerland and Lichtenstein in the process of "reducing their ecological footprint" and optimized their website performance at the same time.

Read more

Hol Dir den Newsletter

Jetzt für unseren Newsletter anmelden und monatlich wichtige Insights aus der Branche und MD Systems erhalten. 

Zur Anmeldung

About MD Systems

MD Systems, headquartered in Zurich, is a unique team of international open source initiative leaders for the Drupal content management system.

With our experts for software architecture and design and our industry solutions, you digitize your organization successfully and efficiently.

MD Systems GmbH

Hermetschloostrasse 77, CH-8048 Zürich

Schweiz

+41 44 500 45 95

[email protected]

  • Contact
  • Impressum
  • Data protection
To top

© Copyright 2023 - 2024 MD Systems GmbH. Alle Rechte vorbehalten. Erstellt mit PRIMER - powered by Drupal.