Wikimedia Interactive Team/Maps Analytics and Design Research

From mediawiki.org

Maps Analytics and Design Research[edit]

Here are some ideas for collecting Maps success and adoption. There a multiple ways to collect data:

  • With Event Logging, a JS API that allows you to send objects of data from the browser.
  • Parsing web requests, and identifying the ones that are map related.

Criterias[edit]

  • With any EL event, differentiate maps embed in articles vs maps in full screen mode.
  • With any EL event, differentiate desktop users vs mobile users.
  • With any EL event, send the number of POIs and elements on the map.

Measuring success rate - KPIs[edit]

GOAL: We want to evaluate if an interactive map is a worthy discovery tool:

  • Are maps a new way to dig the rabbit hole on Wikipedia?
    • Collect maps clickthroughs to Wikipedia articles (1 event per clickthroughs to articles from within a map)
  • How many maps inside articles?
    • we can get it from the tracking category, via api
  • Do users play with our maps?
    • Collect map redraws, i.e. whenever the position/zoom changes (1 event on redraw).
      • I suspect redraws might be too chatty, esp on good hardware.
  • Do users have an interest in maplinks?
    • Collect clicks on maplinks. (1 event when full screen opens)
  • Do users share full screen maps?
    • Send an event when the page opens a full screen map on page load

Understanding user experience[edit]

GOAL: We want to understand how our users interact with the maps

  • Do users click on markers?
    • Collect clicks on markers (differentiate embed in article vs full screen)
    • Collect clicks within marker popups (same as measuring the rabbit hole effect).
  • What is the average session like in full screen mode?
    • Collect dwell time and reflect it to the number of POIs and elements on the map.
  • Desktop traffic vs Mobile traffic

GOAL: We want to evaluate some elements of the user interface:

  • Do users go in full screen mode?
    • Collect clicks on Full screen icon. (1 event when full screen dialog opens)
    • Collect double clicks on map. (1 event when full screen dialog opens)
  • Do users use zoom in / zoom out buttons?
    • Collect clicks on zoom in / zoom out buttons (1 event per click).

GOAL: We want to understand how to improve maps

  • Add a "Give a feedback" button
    • Technical blockers?
  • Make measured A/B tests to introduce/change/remove new features / new controls.
    • Once the service starts to be widely used and changes need to be less disruptive.

Technical questions[edit]

  • How do we send so many events?
    • What do we sample? How do we sample?
    • If collecting them and sending them in bulk, how to we not lose events if leaving the page before events are sent?
      • I think we should send events after one second of inactivity for some events, and for others we might want to reduce how much we collect. Too much data is not very useful either.

See also[edit]