Evaluate a Following-Only Home Feed and Measure Its Impact

A professional social network is considering changing its home feed so that it shows only posts from accounts the member follows. Assume today's ranked feed mixes posts from followed accounts with posts surfaced because someone in the member's network reacted to or commented on them, and with recommended posts from accounts the member does not follow; the proposal removes the last two kinds. What do you think of this idea, and how would you measure its impact? ### Clarifying Questions - Does "follow" include connections (who may be followed automatically), company pages and hashtags, or only people the member explicitly follows? - Would the feed stay ranked, or switch to reverse-chronological order? - Would reshares by followed accounts of posts from accounts the member does not follow still appear? - Is this a change for every member, or an optional feed mode? - What prompted the idea: complaints about irrelevant posts, or a goal such as strengthening ties within members' networks? ### Part 1 — Evaluate the idea Give your view on whether this change is a good idea: what problem it would solve, who would benefit, who would lose, and what you would propose or test instead. ```hint Count the inventory Picture a member who joined last week and follows only a few accounts. What does their feed look like after the change? ``` ```hint The other side of the feed Every post removed from a feed also loses a viewer. Ask who is on the other side of the change. ``` #### What This Part Should Cover - The problem the change would address and the members it would help - Who would be hurt: members with few follows, creators who rely on reach beyond their followers, and the platform's discovery and network-growth loop - A reasoned position and the variants worth testing ### Part 2 — Measure the impact Explain how you would measure the change's impact: the metrics, the experiment, and how you would decide whether to launch. ```hint Shared audiences Randomizing viewers is natural, but then each creator's audience sits in both groups. Ask what that does to creator-side metrics. ``` ```hint Fast and slow effects Some effects appear on the first day; others, such as creators posting less or members forming fewer new connections, take weeks. ``` #### What This Part Should Cover - A primary metric, supporting metrics and guardrails for both viewers and creators - The randomization unit, interference between groups, power and duration - Segment analysis and a launch decision rule ### What a Strong Answer Covers - A clear opinion on the idea, conditioned on evidence rather than asserted - Two-sided reasoning about members as viewers and as creators, including the effect on content supply - Specific metrics with a primary metric, secondary metrics and guardrails, and awareness that ratio metrics can improve only because volume falls - A valid experiment that addresses network interference, novelty and long-term effects - Heterogeneous effects by network size and tenure, and a launch recommendation that follows from them ### Follow-up Questions - Treated members see fewer out-of-network posts, so creators whose audiences sit mostly in control lose little reach. How would you measure the effect on creators without that dilution? - If the change ships as an optional feed mode instead, why is comparing members who turn it on with those who do not misleading, and what would you measure instead? - Suppose sessions per member fall slightly while satisfaction survey scores rise. What would you recommend? - What would you do for new members whose following-only feed would be nearly empty?