Page MenuHomePhabricator

Addition of user interactions with CX suggestions to CX key metrics dashboard
Closed, ResolvedPublic

Description

The Content Translation tool now has custom suggestions menu and various ways for users of CX to customize suggestions, which was developed as a part of T113257.

An initial analysis has been done as part of T381391: Visualize use of suggestions over time for mobile translations.

The team would like to monitor the use of the custom suggestions over time and relevant user interactions for the menu. The goal is to add relevant aggregate metrics to CX key metrics Superset dashboard.

Steps involved:

  • Propose metrics to be monitored
  • Finalize metrics to be monitored
  • Assess if an aggregation data pipeline is needed
  • Develop queries for aggregation
  • Develop Airflow pipeline for aggregation (if needed)
  • Addition of metrics to the dashboard

A new tab has been added to the key metrics dashboard: https://superset.wikimedia.org/superset/dashboard/p/opjO9Nyv5Ry/


Proposed metrics (for monitoring)

  • Daily unique users interacting with the custom suggestions menu
  • Daily clicks to open the custom suggestions menu
  • Aggregate of high level of sources for filters selected (collections topic area, popular articles, previous edits)
  • Aggregate of specific sources filters selected (for example, names of the topic areas, or collections)
  • Proportion of users that proceeded to translation start step by selecting a suggestion from the custom suggestions

to be considered later for funnel analysis

  • Proportion of users (of the ones who opened the menu)
    • that selected at least one filter within the custom suggestion menu
    • that confirmed with at least one filter
    • that closed without any confirmation
  • Proportion of users that successfully published a translation after selecting a suggestion from the custom suggestions

Details

Related Changes in Gerrit:
Related Changes in GitLab:
TitleReferenceAuthorSource BranchDest Branch
analytics_product/cx_suggestions_menu_interactions_daily rework BashOperator as PythonOperatorrepos/data-engineering/airflow-dags!1248brouberolT386675main
Customize query in GitLab

Event Timeline

KCVelaga_WMF moved this task from Incoming to Priority on the LPL Analytics board.

@Pginer-WMF @PWaigi-WMF @SGautam_WMF Please review the proposed metrics and let me if any changes are required. I will start working on the data pipelines after we finalize the metrics.

Thanks @KCVelaga_WMF for sharing this, I have a couple of questions here to clarify.

  1. What do we mean by "Daily unique users interaction..."?
  2. We also have search option available when users land on custom suggestions screen. @Pginer-WMF I am wondering if it's worth checking? If so, this can be part of following point - "that selected at least one filter within the custom suggestion menu" - How they selected it by long scroll or using the search option.
  1. We also have search option available when users land on custom suggestions screen. @Pginer-WMF I am wondering if it's worth checking? If so, this can be part of following point - "that selected at least one filter within the custom suggestion menu" - How they selected it by long scroll or using the search option.

From my perspective I think about these metrics in two groups: one group of metrics is more directly related to impact (users starting/making translations based on the filters we provide), and the other group is more about understanding the user workflows. For the latter, I think it would be great to capture them as a workflow funnel so that we can better understand which are the drop-off points (including search) as we did with the entry point analysis.

@SGautam_WMF

What do we mean by "Daily unique users interaction..."?

Ah the original wording is a bit confusing, corrected it. It is number of daily unique users interacting with the menu. We have a metric for absolute number of clicks to the menu, but a single user might be clicking it twice or more. So this is how many unique users.


@Pginer-WMF That's a great point. I think Sudhanshu's second question, and all the metrics mentioned as "Proportion of users" are better suited for funnel analysis. I divided the metrics into groups to simplify.

@KCVelaga_WMF The 1st set captures all the metrics we need on the dashboard, and having the 2nd set is useful for funnel analysis to guide more feature improvements.

  • Some of these metrics will be useful for campaign organizers who agreed to add their collections to the dashboard. We hope to share this with them at the end of Q4.

Note: A data pipeline needs to be setup for the necessary aggregations. Although the data is small enough for direct queries for now, in future, as the volume grows we will most likely run into timeout error on Superset. I will start working on the queries for the pipeline this week.

Change #1135388 had a related patch set uploaded (by Brouberol; author: Brouberol):

[operations/deployment-charts@master] airflow-platform-eng: grant task pods egress permissions to gitlab

https://gerrit.wikimedia.org/r/1135388

Change #1135388 merged by Brouberol:

[operations/deployment-charts@master] airflow-analytics-product: grant task pods egress permissions to gitlab

https://gerrit.wikimedia.org/r/1135388

brouberol opened https://gitlab.wikimedia.org/repos/data-engineering/airflow-dags/-/merge_requests/1248

analytics_product/cx_suggestions_menu_interactions_daily rework BashOperator as PythonOperator

kcvelaga merged https://gitlab.wikimedia.org/repos/data-engineering/airflow-dags/-/merge_requests/1248

analytics_product/cx_suggestions_menu_interactions_daily rework BashOperator as PythonOperator

KCVelaga_WMF moved this task from Priority to Review/sign-off on the LPL Analytics board.

A new tab has been added to the dashboard to track the required metrics: https://superset.wikimedia.org/superset/dashboard/p/opjO9Nyv5Ry/

Please explore and let me know if you have any questions/suggestions.

KCVelaga_WMF changed the task status from Open to In Progress.Apr 18 2025, 1:28 PM
KCVelaga_WMF edited projects, added: Product-Analytics (Kanban); removed: Product-Analytics.