Page MenuHomePhabricator

[Design] Create design for empty state of search experiment
Closed, ResolvedPublicDesign

Assigned To
Authored By
ovasileva
Jul 29 2024, 10:13 AM
Referenced Files
F57150147: image.png
Aug 7 2024, 1:29 PM
F57149998: image.png
Aug 7 2024, 1:29 PM
F57148582: image.png
Aug 7 2024, 1:29 PM
F57148501: image.png
Aug 7 2024, 1:29 PM
F57148468: image.png
Aug 7 2024, 1:29 PM
F57146833: image.png
Aug 7 2024, 1:29 PM

Description

Background

  • We would like to run an experiment on the UI of the empty state of search. This ticket will cover creating the designs for the experiment

User story

  • As a developer, I would like to know what the best way to deliver an experimental version of the empty search state is so that I can understand whether the empty state of search is an effective placement for recommendations

Requirements

Acceptance criteria

  • Create designs for empty state of search experiment
  • Define language for testing (with @ovasileva and @sgrabarczuk )

Event Timeline

ovasileva created this task.

I just had a great conversation with @dchen about this. Here is the recommendation to get directional data that answers the question: "Is the empty state of Search an effective location to show recommendations?"

We handpick a series of articles and launch a survey of some kind on those articles. We could also do all articles, or a subset based on some logic. Maybe trending articles?
In the survey, we show mockups of 3-5 different places where we could show recommendations on those pages. The mockups don't even have to show the content of the recommendations themselves. We could redact them somehow or use placeholder text.
We ask participants either to pick their top preference, or their top 2, or however many. We could also ask them to rank the different locations:

  1. Empty state of search
  2. Footer like in Minerva
  3. Appended to an infobox
  4. Sidebar like where the appearance menu is
  5. Maybe in the hatnote location above the introduction paragraph
  6. And we can think up a few more.

Nice to have would be for the order of the locations to be randomized for each participant to avoid cognitive bias.

This would give us a sense of the stated reader preference for different locations. It won't give us what they would actually do in a real reading session, or whether they would click on recommendations at all. But if we are going to show recommendations somewhere, it could give us the location that would have the highest yield.

JScherer-WMF added a subscriber: Jdrewniak.

Needs to be assessed for technical feasibility. Had a Slack conversation with @Jdrewniak about it and it seems possible, but we need next steps.

Some next steps that we could do with this. I'll assume for now that it'll be a static survey on the project page somewhere with screenshots and a multiple choice question.

  • Mockups for each recommendation option as GIFs ideally...
    • Empty state of search in Vector
    • Footer of Vector
    • Bottom of an infobox
    • Right sidebar of Vector
    • Left sidebar of Vector
    • contentSub location
    • What else? - ask in design review on Thursday
  • Mockup the mediawiki survey in my sandbox.
Jdlrobson subscribed.

We need some input from Olga on this ticket so moving to next sprint.