PM Quest · Practice cases · Learning resources

Guardrail Metrics for Facebook's Engagement Optimization

Metrics Definition · beginner

Facebook is focusing on optimizing for increased user engagement, specifically aiming to boost the average time spent per session on their platform. While this metric is believed to drive advertising revenue, there's a concern about potential negative impacts on user well-being. Identify which guardrail metrics Facebook should track to prevent diminishing user satisfaction or increasing negative content exposure. Avoid suggesting changes to content algorithms yet. Prioritize the most critical guardrail metrics that Facebook should monitor.

Write your approach on PM Quest

Community approaches

Yuvraj’s approach

Identify key guardrail metrics to track: Negative feedback rate: the frequency with which users take action to hide posts, unfollow accounts, or report content. Use Sentiment score: Data gathered from periodic in app surveys asking users how they feel about time spent on platform. Active VS Passive Consumption Ratio: The ration of meaningful engagement-comments, share, like, message, etc and passive scrolling.

Justify why each metric is essential: Negative feedback rate directly signals that the algorithm is serving material that users find irrelevant, offensive or harmful. Sentiment score captures emotional impact that raw engagement metrics miss. High passive scrolling consumption is often related with lower long term satisfaction.

Assess potential tradeoffs between engagement and user satisfaction: Optimising purely for time spent can inadvertently reward sensational, devisive or addictive content. While this boosts short term ad revenue, it might lead to long term user fatigue and churn. Balancing these requires accepting a potential slight decrease in daily active minutes in exchange for a healthier and more sustainable user base.

Recommend a primary guardrail metric to prioritize: The primary metric to prioritise is the negative feedback rate. Unlike survey data which can be delayed, this metric provides real time actionable feedback. If an increase in average session length correlates with spikes in negative feedback, the algorithm can be adjusted immediately before more dissatisfaction occurs.