Page MenuHomePhabricator

Instrumentation for Search Improvements A/B test & Nav Pt 2
Closed, ResolvedPublic

Description

Background

This task is intended to capture what data wiring is needed to answer if this feature met its goals.

Measurement Goals [This will be updated pending discussion of revised Measurement Plan]

Validation
These will be measured via an A/B test

  • KR 1.1 8% increase in users initiating Search from article view
    • Control group: unique users in A who have initiated at least one search from article view/unique users in A who have had article views
    • Variant group: unique users in B who have initiated at least one search from article view/unique users in B who have had article views
  • [Primary metric] KR 1.2 1% higher app retention rate for logged-out users that had the more prominent article search (using 2-week retention)
    • Control cohort: of the unique users in A who engaged with the control search icon, what % were retained to the app?
    • Variant cohort: of the unique users in B who engaged with the variant search bar, what % were retained to the app?
  • KR 1.3 5% increase in average pageviews/unique user for those who engaged new article search bar
    • Control group: average pageviews/unique user for users in group A who initiated at least one search from article view
    • Variant group: average pageviews/unique user for users in group B who initiated at least one search from article view

Guardrails

  • GR 2.1 No more than 10 negative user complaints about changes to the Navigation that can’t be addressed
  • GR 2.2 No more than 5% increase in app uninstalls (vs baseline)
  • GR 2.3 No more than a 10% increase in app crashes

Curiosities

  • CR 3.1 Did account creations increase as a result of the Profile menu being accessible from all screens? (Baseline of 15 days before vs Experiment Date Range)
  • CR 3.2 Do users prefer a different location for the article view search?
  • CR 3.3 What was the feature retention rate for article search for both A & B? (Recalculate 1.2 but using return to engage with article search as definition of retention, not return to overall app) Please split by Logged-in and logged-out
Link to Finalized Designs

image.png (1×1 px, 483 KB)

    1. Tasks
  • @ PM Add measurement goals to task
  • @ PM Add link to final designs
  • @Analyst Create schema Doc
  • @analyst Create validation task after Beta release

Event Timeline

@SNowick_WMF could you please begin on this week, specifically what you will need for the A/B test (to unblock Engineers assigning users : T383829: Assign users for Article search bar A/B test

Clarifying how we will split data since it's relevant for article view data needed for our test.

The Search engagement events (and test group designation) will be tracked in schema ios_search - we may/may not need to update this instrumentation after re-design, search event starts are measured by source values. This will all be documented in Instrumentation docs.

We will use the article views by test group users using impression event data from ios_article_link_interaction.

Noting here that we will likely need to decide whether to use the article link data being implemented from task https://phabricator.wikimedia.org/T374036 to track Search engagement by group for KR1 or build a separate similar tracking instrumentation so we can get insight into Search user article views. Will map this out in my Instrumentation docs but want to link these 2 tasks for tracking later if needed.