Analyze homepage drop and feed ranking

You are being considered for a product data science position at LinkedIn. Respond to the two product-sense questions below.

1. Diagnose a drop in Home Page → Profile Page traffic

LinkedIn observes that weekly volume from Home Page to Profile Page has dropped substantially. Your task is to decide whether this reflects a product defect, a measurement or logging problem, a shift in traffic composition, or an intentional product improvement.

Explain how you would investigate the decline. Include:

• your definition of the primary metric and its denominator
• what sanity checks you would run for bugs, instrumentation changes, and releases
• which macro metrics you would examine first
• how you would trace user paths after they arrive on the Home Page
• how you would break down the analysis by platform, geography, member cohort, and traffic source
• how you would interpret a scenario in which overall session duration stays flat even as Home → Profile visits decline

Assume LinkedIn just shipped an inline profile preview on the Home Page: hovering over a member's name or card reveals essential profile details without opening the full Profile Page.

2. Measure the success of switching the feed default from "All" to "Relevant only"

LinkedIn intends to change the Home Feed default so that, instead of all content, viewers see only the content judged most relevant to them.

Design an experiment and a success framework for this change. Include:

• the central product hypothesis
• the experiment design and randomization unit
• primary success metrics for viewers
• poster-side metrics and possible creator concerns
• guardrail metrics
• short-term versus long-term tradeoffs
• how you would read a result where viewer engagement rate improves but some posters get lower reach

Be explicit about metric tradeoffs, possible selection bias or traffic-mix confounding, and your decision rule for whether the launch succeeds.

Overview: This question tests product analytics, causal reasoning, instrumentation and logging checks, segmentation and cohort analysis, and experiment design for feed ranking and user-path behavior.