Improving user journeys on GOV.UK
In 2018, a user couldn't move easily from one part of GOV.UK to another. The website was made up of three separate and siloed areas, with no reliable way of navigating across them.
Users couldn’t find what they needed or complete tasks, and publishers couldn’t easily link content together.
Having three websites masquerading as one is a problem.
The main navigation, page layouts, related links, and breadcrumbs all worked differently depending on you where you were.
There were two separate topic systems, and no way to tag a piece of content to both. Related links needed to be curated by publishers, but even they didn’t know how to navigate the two taxonomies to find the right pages.
Most user journeys on GOV.UK begin on a content page. This means that related links — not navigation pages — are one of the best ways to help users find what they need.
Unfortunately, the vast majority of content on the site had no related links at all. We wanted to find a way to use machine learning to add links to pages so that publishers weren't overwhelmed with an impossible task.
What we did
We realised early on that to fix navigation, we needed to fix the foundations of how GOV.UK fit together. We needed a robust and scaleable information architecture.
We tackled the taxonomy
The team created a new, overarching taxonomy to sit across the different parts of the site, unifying all content. This would help us link the separate areas of the site together to improve user journeys.
We mapped the domain
Mapping the domain meant bringing together all the parts of the website and exposing their relationships. For example, a ‘Minister’ can give a ‘Speech’ on behalf of a ‘Department’.
Once we had an overview of the domain, we needed a way to organise and prioritise things in a way that would be easy for users to understand.
We organised content for our users
Users have different needs depending on what they are doing. By defining the ‘jobs’ and 'tasks' our users do, we were able to figure out what type of content they needed, and when.
Each user job corresponds to a different type of content. We ran a series of card sorting sessions and tree tests with publishers and end-users to map content types to jobs. This gave us insight into which pages needed to link together.
We used machine learning
We understood how the site fits together, and we had a rich, deep taxonomy. Jobs gave us a way to understand journeys and prioritise features. We used these elements with machine learning to generate related links.
Publishers could choose from a list of machine-generated related links or add their own. We could now add related links to the thousands of pages that had none.
The work done to streamline the front-end templates meant that we were able to ship and AB test our related link formulas with ease.
A range of navigation improvements followed the related link work. The biggest was a new, global menu that helps users to understand what GOV.UK offers and find what they need. A range of new browse pages have also been shipped.
The unified taxonomy is being iterated and deployed continuously as we merge and retire the two older taxonomies. Tagging is simpler for publishers, and tags can be updated easily when the need arises.
This foundational work provides the basis for structured content, publishing improvements and better navigation.