Page MenuHomePhabricator

[Spike] How do people add images in VisualEditor
Closed, ResolvedPublic

Assigned To
Authored By
AUgolnikova-WMF
Apr 9 2025, 2:44 PM
Referenced Files
F59290371: media-dialog-percentage.png
Apr 18 2025, 3:04 PM
F59290004: window-open-from-tool.png
Apr 18 2025, 2:17 PM
F59290002: window-open-from-context.png
Apr 18 2025, 2:17 PM
F59290000: window-open-from-command.png
Apr 18 2025, 2:17 PM
F59289997: dialog-done.png
Apr 18 2025, 2:17 PM
F59289991: dialog-abort.png
Apr 18 2025, 2:17 PM
F59289986: dialog-insert.png
Apr 18 2025, 2:17 PM
F59290035: cross-wiki-uploads.png
Apr 18 2025, 2:17 PM

Description

As part of planning for FY25-26 under WE2.1 KR, we are exploring an idea to suggesting relevant image improve image discoverability in the existing VE flow on Web. Our hypothesis is that this will increase image contributions on Wikipedia and decrease uploads to Commons which have shown to have a significantly higher deletion request rate.

We want to understand people's intentions of editing an article and adding images.

We have captured the user flow and problems during the research phase based on existing research on Miro

Questions:

  • [Research - outside of scope of this ticket, just as FYI] Do people who are already in the editing flow of VE have an intention to add an image? Have they already found an image before starting to edit?
  • [Data] How many people/sessions use media tab in VE for finding or uploading images? Quantify interactions in VE with the media tab: usage of mediasearch, upload tab, search on commons (if we can connect the session), mobile/desktop

As an outcome, we want to know

  • What data do we track in VE from above? then determine next steps. Get the data if easy

Event Timeline

AUgolnikova-WMF renamed this task from How do people add images in VisualEditor to [Spike] How do people add images in VisualEditor.Apr 14 2025, 4:25 PM
AUgolnikova-WMF updated the task description. (Show Details)

There is an existing dashboard (https://superset.wikimedia.org/superset/dashboard/222/, built for T259308 & T260254) with a bunch of relevant metrics, although it looks like it is no longer producing up-to-date data. @nettrom_WMF is going to check whether it would take a lot of effort to resurrect.


It looks like the VE media dialog is instrumented quite extensively. Below is a list of all interactions that are currently instrumented.

See doc here: https://docs.google.com/spreadsheets/d/1TJM51_xsfvRsnC_gvQ6GU8Cy_4lcclddev88UtO3924/, or screenshots below:

search-change-query

Changed search query (every keypress)

search-change-query.png (373×602 px, 44 KB)

search-clear-query

Cleared search query by clicking on indicator

search-clear-query.png (372×602 px, 48 KB)

search-choose-image

Selected image from the search results

search-choose-image.png (373×602 px, 40 KB)

search-confirm-image

Confirmed selected image by clicking "Use this image" (which then takes user to the captions step)

search-confirm-image.png (372×602 px, 42 KB)

search-change-image

Arrived back at search, either through hitting the back button from a selected image, or by clicking "Change image" from the captions step.

search-change-image.png (372×602 px, 43 KB)

search-upload-image

Initiated file upload

search-upload-image.png (373×602 px, 47 KB)

NOTE: it appears that only about half of these end up resulting in an actual cross-wiki upload. Below are the amount of files on Commons tagged as cross-wiki uploads:

cross-wiki-uploads.png (373×602 px, 41 KB)

dialog-insert

Clicked "Insert" to add selected image to the page

dialog-insert.png (373×602 px, 42 KB)

dialog-abort

Closed media dialog without making changes

dialog-abort.png (373×602 px, 41 KB)

dialog-done

Clicked "Apply changes" to edit existing image on the page

dialog-done.png (373×602 px, 40 KB)

window-open-from-command

Opened media dialog by doubleclicking an existing image

window-open-from-command.png (372×602 px, 44 KB)

window-open-from-context

Opened media dialog by selecting "Edit" from the context menu that appears when clicking an existing image

window-open-from-context.png (373×602 px, 43 KB)

window-open-from-tool

Opened media dialog through the toolbar

window-open-from-tool.png (373×602 px, 42 KB)


I'm sure not all of it will be useful, but that's pretty much what is available, with data going back 3 months (unless we actively start storing these).

Oh I forgot about mobile/desktop. If anyone needs any of the above actions broken down by platform, LMK, but until then, here's a breakdown of all of the above interactions in total, by platform:

desktop: 3582875
phone: 268363

Another one: the amount of times the media dialog is opened as a proportion of the total amount of initialized VE sessions:

media-dialog-percentage.png (373×602 px, 40 KB)

There is an existing dashboard (https://superset.wikimedia.org/superset/dashboard/222/, built for T259308 & T260254) with a bunch of relevant metrics, although it looks like it is no longer producing up-to-date data. @nettrom_WMF is going to check whether it would take a lot of effort to resurrect.

From @nettrom_WMF:

The notebook that populates the underlying data appears to run just fine, so I'll set it up to update daily moving forward. I'll also backfill it with data going back to Feb 1. If more work is needed (e.g. the desktop/mobile split mentioned in T391488), I think that's a good candidate for a consultation hour. If team needs this more permanently, then an R&DS request is needed.

I've completed backfilling the data from Feb 1 onwards. There's now a daily cron job that updates the underlying data, it runs at 05:20 UTC every day. The dashboard has been updated with filters so that it by default shows the last quarter's data. Older data is still available, it's just not shown immediately.

@nettrom_WMF thanks a lot for reviving the dashboard. I am trying to get older data - 1-2 years, but when I set the timeframe to last year only the last quarter shows.

@AUgolnikova-WMF : There's a gap in the dashboard between Dec 20222 and Feb 2025 because the dashboard was no longer in use, and because the underlying data source doesn't go back further than 90 days. The "Daily Media Searches in VE" chart is different, it'll always show the last 90 days because that's querying event data directly (and that's the same data source that used for the aggregations shown in all the other charts).

I hope that clarifies, although it doesn't make it any less unfortunate that we don't have that data.