Sprint 11

1 Nov 2017 to 14 Nov 2017

Done

Learnt

  • New table prototypes tested well with users
  • Including a table of contents on the dataset landing page was a great way to give an "at a glance" view of the metadata available
  • The show/hide component is sometimes missed, because it's not that obvious that it's interactive
  • Time filter content needs to be more contextual e.g. "Add a single month" not "Add a single time"
  • Download file is testing well, but it needs some simple contextual information, for example; dataset name, filters added, date customised

Sprint 10

19 October 2017 to 31 October 2017

Done

Learnt

  • Users want to filter dates in this order of priority: latest date, single date, date range, custom mix of dates
  • Navigating between release years causes big problems for users
  • Users expect to be able to use breadcrumbs to go up a level to pick other release years
  • Version history labelling could be improved (e.g. “original” and “revised”)
  • Many users don't notice the "shopping basket" component
  • Users are naturally drawn to maps but expect to be able to mix and match between different ways of interacting with them

Sprint 9

4 October 2017 to 17 October 2017

Done

Learnt

  • Floating table of contents not being noticed
  • User thought horizontal black bar below page intro was the end of the page
  • Chart tooltip content is only available in the tooltip (it's easy to miss and not very accessible)
  • 2 of 3 users preferred larger font size because it didn't strain their eyesight
  • 2 of 3 users expected contact details to be at top of page
  • Users confused when asked to add 2nd date range
  • Users expect long lists to be grouped or structure somehow
  • Users expect a link to other similar datasets from dataset landing page
  • Users not keen on ow.ly links from social - they don't know what they are going to get
  • If tweets cover difficult topics (e.g. abuse), users prefer no illustration (text alone is fine)
  • Reordering content can give massive improvements on click throughs
  • Some screen reader users find heading essential for navigation
  • Buttons should be marked up as buttons, not p tags
  • Labels and headings should follow the pattern of "plain english name, code" for example "Food K02000001"

Sprint 8

21 September 2017 to 3 October 2017

Done

Learnt

  • Skip links need to send focus to the correct area
  • Every interactive element needs a visual focus style
  • All interactive elements need clear, consistent active, focus and hover styles
  • Non-interactive elements shouldn't be focusable
  • Links need to be unique and descriptive
  • Colour contrast on charts, links and text needs to be improved
  • Chart tooltips are only accessible via mouse (not keyboard)
  • Areas need context: e.g. “Clifton (electoral ward) in Bristol”

Sprint 7

7 September 2017 to 20 September 2017

Done

  • Closer integration of the web front-end and the API
  • Realtime generation of a filtered CSV through the web front-end
  • JavaScript enhancements to the cart so it updates instantly, without refreshing the page or additional action
  • Improved the date filter so that the end date is updated based on the start date chosen by the user
  • Added the "apply filters to latest dataset" alert
  • Enabled the entire row to be clickable on the filter options page
  • Design Sprint to tackle "find data by area" (Google Slides)
  • Design Sprint prototype
  • Built a backend events reporter to catch errors

Learnt

  • If users see a map, they expect it to be interactive
  • Most users get confused when shown multiple statistical areas for one area (for example, Cardiff Electoral Ward and Cardiff Civil Parish)
  • Users expect to see data at all levels of the area navigation journey

Sprint 6

23 August to 6 September 2017

Done

Learnt

  • After fleshing out the types of data we are storing in postgres, it made more sense to start moving our database to a mongodb datastore. This allowed us to store more complex data types and have quicker database reads
  • Versions and instances share a lot of data, but we interact with them in 2 separate APIs and store them twice. Moving instances into the dataset API will prevent duplication of data and time spent copying it over.

Sprint 5

9 August to 23 August 2017

Done

Learnt

  • When selecting a range, users expected the end date to pre-populated with start date
  • Users expect to wait for downloads to generate, but need to be kept informed
  • Links should be more descriptive, for example 'view 4' should say 'view 4 regions'
  • Repeated headings and links like "download this table" need to be unique and descriptive
  • Users need to know what's changed between dataset editions
  • Charts need to have useful descriptions (alternative text) or be available in tabular form
  • Table previews need to show all columns, but only a few rows
  • Functionality should be clear from the way something looks, for example a drop down menu needs to have a arrow that acts as a visual cue
  • Consistency is very important for accessibility

Sprint 4

26 July to 8 August 2017

Done

Learnt

  • Acronyms cause confusion, users don't understand that UA means Unitary Authority
  • Including the word 'save' on the primary button makes users feel more confident even if auto-save happens in the background
  • Labelling needs to be clear when search boxes don't search the entire site

Sprint 3

July to 25 July 2017

Done

Learnt

  • Andy can make decent cookies
  • Mobile traffic is as at 23%, tablet traffic is at 7%
  • iOS autocorrect does a bad job of autocorrecting acronyms that users type into search (like GDP, RPI and QMI)
  • The floating table of contents is causing multiple issues
  • Long legend text on time series charts are being cut off on smaller screens
  • Zip files can't always be downloaded on iOS

Sprint 2

28 June to 11 July 2017

Done

Learnt

  • Users find the breadcrumbs useful for navigating around the process
  • Users need a table preview so they can check their filters
  • Users need to know filesize before they download a spreadsheet
  • Users need instant confirmation of what filters have been applied (also known as a basket)

Sprint 1

14 June to 27 June 2017

Done

Sprint 0

7 June to 13 June 2017

Done

Data Discovery Alpha

See what we did during the Data Discovery Alpha.