Designing a knowledge taxonomy should be simpler than you think

When our Knowledge Repository first went live in 2017, it was as a free-standing website, the Major Projects Knowledge Hub, and we borrowed the subject taxonomy and the other elements of the tagging from the Crossrail Learning Legacy (with their kind permission). Then in 2019 we integrated the repository into the new Association website, which gives members a one-stop shop for everything from events to professional development to knowledge. And we were careful to retain the open access nature of the repository to benefit both members and the wider community.

We’ve recently embarked on the first of a two part set of activities to refresh the repository, using what we have learned over the last 6 years as well as some research we undertook with a number of members. The second part of this project will see a major refresh of the landing page and the general design of the repository. But first of all, we are replacing the original Crossrail Learning Legacy taxonomy and the associated metatags with a brand new taxonomy.

What’s the basis for the new taxonomy?

The new taxonomy has drawn on a number of sources, including project bodies of knowledge and the main intention behind it has been two-fold:

1. We want to reflect the actual make-up of the c.2000 resources that the repository contains, which is a mix of Association reports, learning legacy content from (London 2012, Crossrail, Thameslink and HS2), consultancy, industry and Government reports and guidance and research;
2. We want to make the search and filtering more intuitive, simpler and give you a better quality of search results.

Check it out

So far, we have recoded around 15% of the repository content, so do feel free to check out the work-in-progress. If you use any of the three filters, topics, content types or sectors, double check that the results you generate seem to chime with what you would expect. We will be removing the filters intended audience, knowledge types and format, because these are either problematic from the perspective of adding new content or largely already provided for in the top three filters. The filters currently display the taxonomy elements on the basis of ‘most-used’. We will be changing this to alphabetical order because that makes scanning down the list far easier.

We will also be removing the project life-cycle tags: initiation, implementation, handover and operations, not because these aren’t important but rather because it is impossible to assign the vast majority of the repository content to any one of these stages in the life-cycle. We have retained them in the topics, which means we can still tag those few resources that can be collated on this basis.

We have also dramatically simplified the subject taxonomy and to replace the original three levels of sub-coding with a flat single level of topics.

Lessons for other knowledge repositories

What have we learned over the past six years and how are we using it to inform our refresh:

1. Beware the distinction between how you tag knowledge and how users search for it
   Bodies of knowledge and technical taxonomies work OK for tagging explicit knowledge (i.e. if someone wants to know how to line a tunnel with cement or how to structure a business case) but when it comes to tacit knowledge, they are far less useful. We have found in the majority of cases the way in which resources are tagged, which is entirely logical from the perspective of the person doing the tagging, rarely reflects how users search for knowledge;
2. Don’t be tempted to overcomplicate your subject taxonomy, it doesn’t make it more precise
   The original Knowledge Repository taxonomy tagged content down to three levels of subject. No technical taxonomies are ever entirely intuitive and multiple levels leave users trying to guess under which top-level subject they will find the tags they were looking for. Which is hardly a great user-experience. If your keyword search is up to scratch this may be a far better way of finding arcane content. This, if prefabrication of modular components is what the user really wants to find out about, then the best way of ensuring they find what they need is by ensuring the resource title or the abstract is precise in it’s use of language.
3. Don’t forget: most people know they need some help but that doesn’t mean they can articulate their search
   It’s all very well having a sophisticated set of subject tags but that may only work for the one in 250 people who know exactly what they are looking for. Most of us may struggle to articulate whether the problem we are trying to address is related to governance or to assurance, or is simply a function of poor project organisation! For many users, the search and filtering process acts both as a means of finding an answer and as a way of trying to diagnose what the problem is in the first place;
4. Serendipity and personal curiosity may be the most effective driver of repository usage
   Everything we have learned from Facebook, LinkedIn, Twitter, TikTok or Instagram reinforces the fact that the behaviour of digital natives is almost entirely driven by a tiny attention span. Thus, catching someone’s interest with an image or a resource title and then using related materials to take them further into a collection is likely to be one of your most effective knowledge sharing strategies. Knowledge acquisition and sharing is never ordered and structured, even if our repositories like to pretend that is the case;
5.  Most people know what they need when they see it
   Content curation is a very underrated activity. We should learn from the media streaming sites, Netflix, Disney or Amazon Prime, who understand the value of curating Christmas collections in December or of offering Scandi Noir films and series for the fans of that genre. This final point anticipates the second stage of our refresh project, when we will start to use semi-permanent and temporary collated collections to populate the landing page of the knowledge repository. The semi-permanent collections will reflect the actual make up of the repository content (for example, providing a collection based on learning legacy content) and the temporary collections will allow us to generate time-limited collections to mirror what’s going on with the Association’s event programme; encouraging those who are fired up by attendance at a full-day seminar to pursue that interest further in the repository.