I aligned on scope: LinkedIn core product only, excluding Learning. Engagement rate—average session time per logged-in user—has dropped 10% over six months. No measurement or pipeline changes. The decline is global, consistent across demographics and platforms, and first-time seen (not seasonal).
The pattern shows that total users and session starts are stable, but session time is falling mainly on two surfaces: the feed and the jobs page. That rules out login or access friction and points toward in-session experience degradation.
External factors were ruled out—competitors like Indeed or Monster are unaffected, no regulatory or PR issues, and no market downturns. So the cause is internal to LinkedIn.
Investigating internal levers, the main recent change was an update to the recommendation system that increased the ratio of sponsored content in both feed and jobs listings. As ads and promoted posts rose, organic content visibility dropped. The algorithm thus reduced relevance: users encountered repetitive or low-value influencer and ad content, spent less time scrolling, and exited faster. Engagement loss appeared across all segments because both professionals and job seekers were exposed to the same higher-ad mix.
Root cause: excessive prioritization of sponsored content within the recommendation algorithm, lowering content relevance and perceived professional value, leading to shorter sessions and fewer interactions.
Immediate actions:
Roll back ad-weighting parameters to restore the prior organic-to-sponsored ratio.
Add a visible “show me more relevant content” or “hide similar ads” feedback option that directly trains the recommender in real time.
Cap consecutive sponsored posts in feed and job listings.
Run A/B tests across cohorts to find the optimal ad density that balances revenue and engagement.
Communicate transparently with users about improving feed quality to restore trust.
Long-term actions:
Refine the ad ranking algorithm to include a “career relevance” signal so promoted posts must align with user industry and goals.
Introduce explicit user interest selection (“choose your top 3 focus areas”) to personalize feed content.
Apply adaptive ad frequency logic—more engaged users see fewer ads (inverse Fibonacci pattern: 1,2,3,5,8…).
Expand relevance audits combining engagement telemetry, dwell time, and feedback sentiment to keep the recommendation system tuned.
Metrics to measure success:
North star: average session duration per active user (target recovery +5–10% vs baseline).
L1 metrics: scroll depth, content interaction rate, ad skip/dislike rate, and CTR on job listings.
Counter metrics: sponsored post impressions and ad revenue stability within acceptable range.
Summary: the engagement drop stems from algorithmic over-optimization for monetization—too many or poorly targeted sponsored posts diluted professional relevance. The fix is to rebalance the feed, integrate user feedback loops, refine ad relevance, and dynamically adjust ad frequency so content remains valuable while maintaining ad revenue.