Determinants of Social Media View Growth: Evidence from Channel-Level and Post-Level Analysis

This study examines the drivers of social media performance across both large and moderate following accounts. We combine channel-level panel regressions with post-level predictive benchmarking, drawing on top-account snapshots from TikTok and YouTube collected in June, September, and December, alongside a Facebook posts dataset from an account averaging around 200 interactions per post. This allows us to estimate how engagement and audience size relate to view counts, and to compare the predictive accuracy of five model families. At the channel level, likes growth is the strongest predictor of view growth (elasticity ≈ 0.75–0.86), followed by subscriber growth (≈ 0.15–0.20) and comments (≈ 0.05–0.15). These results hold across multiple robustness checks. At the post level, XGBoost and Neural Network models perform best, and feature importance analysis shows that distribution-related variables, such as page size and engagement, account for most predictive power, while content and timing features provide smaller but consistent improvements. For moderate-following accounts in particular, audience size proves far more influential than any content variable. Overall, distribution appears to play a larger role than content design in driving near-term social media performance, though content optimization still provides incremental gains. We conclude by discussing implications for creator strategy, robustness considerations, and data-access limitations.