Designing with data

Following the GDS service standard we wanted to make sure we were using data on the alpha prototype wherever possible to help us make decisions. We were also keen to put tools in place to help us analyse the success of the alpha and allow us to make more decisions in the future based on real data.

Our main source of information was analytics from the current website, however the structure of the current site and implementation of our analytics software means that we are wary of relying on some figures too much. We have been unable to change or even keep the tags up to date and as a result have some concerns with using some of the more in-depth features. There have been however a number of places on the alpha where we were able to utilise this data to good effect.

Using data in the alpha

Our Analytics Team did a great bit of work looking into the relative popularity of content and allowing us to provide a more logical ordering to our taxonomy. We had decided the main four themes should stay alphabetically but wanted to try to order the menus within these by importance. We were keen however that if we were going to deviate from a simple A-Z that we needed to base this on solid information.

The team looked into the relative popularity of content under each area and were able from this to come up with a more sensible ordering based on what users had shown to be more important to them. For example our people, population and community drop down.


We were also able to use the analytics intensively around search. Knowing what users are searching for is essential to building a good search, and we used these terms in testing the newly developed Elasticsearch. We could have done more in this area but were held back slightly by the three-month time scale of the alpha and developing search with only limited content available until the later sprints.

One of the key elements of our search design was having ‘sticky’ items for key search results. It is an idea that we have wanted to try for a while and the alpha gave us this opportunity. We knew which pages we would want to direct users to in the first instance (this had been a consideration when they were designed) and used the analytics to identify which terms would direct to which pages. Although in the end we went with a technical solution using the inbuilt relevancy to highlight these pages this was great as a comparison and to ensure we were getting the results we expected.

We were also able to dig into the analytics and identify key synonyms between user terminology and the more formal statistical language. Whilst we tried to simplify this language wherever possible the nature of the content means there will always be a need for some technical language and this was a useful way to bridge that gap.

Perhaps most interestingly though this formed the basis for populating the look-up table generating the call to action in search results when ONS do not provide the data and allowed us to direct users to the department that would most likely to have that data. This is a feature we think could be really useful for users and something we will no doubt look to build on in the future.


Looking to the future

To enhance what we can do in this area in the future we have supplemented our analytics data on the alpha with a search console that logs user searches and displays if there were any results or not. The no results are of particular interest as we are looking to simplify search and this will be something we will need to watch closely as we go forward and feed back into search changes or perhaps additional synonyms.


We have also decided to implement a vanilla version of Google Analytics rather than use Web Trends as we do on the current site.  This is a good opportunity to try out another tool and should give us a much better idea of what we want from our analytics in the future.

The analytics,  structure of the alpha and additional tools should also give us much more opportunity to use data and we will continue to use it as much as possible in the future particularly if we get approval to move to the beta phase.