The challenge with the ONS dataset management was to enable the user to use a tool as a way to manipulate huge datasets (10,000's of fields) online, the user needed to be able to view the data online as well as download the raw data created. The dataset model was defined by geography and subject matter or topic - to create a dataset you basically needed to define "what & where"
Complex geographies
the complexity lay in the geographical hierarchy and the fact the user could select different sub levels in different areas within the UK and then slice that up by a topic of interest
workshops were held with stakeholders and users alike to understand how "geography" works
as well as applying a user centred approach, a great deal of time was spent understanding the complexities of the geographies from SME's and users alike
as well as being part of the datasets structure these geographies were to be mapped out visually to assist the user, they may not know the name of a place but roughly where the area is and vice versa
Tricky topics
The datasets had rules around them where the topic fields could only be valid in conjunction with other fields, in essence we needed a set of business rules that would satisfy a "if" "or" matrix model. This was down to the datasets being created as almost a 3 dimensional field, with your X,Y & Z fields being represented by topc, topic and geography - even this description doesnt do the complexity justice.
this is an example of a topic data table without the 3rd dimension - this would normally be a tab/field on a spreadsheet that you would click to add once a selection had been made
this shows a working example of how multiple topics are mandatory until paired topics are selected, clicking one may or may not mean that you need to add another. this can be seen in the pdf here
once the topics required have been selected you can then select the elements of that topic you wish to use
the user is finally presented with an explanation of their dataset and options to view or download it
the prototype shown here underwent continous testing - sample of which can be seen here