Predictors of User Satisfaction in Personalized Algorithms: An Empirical Analysis of Control, Accuracy, and Relevance Across Media Platforms
While digital platforms invest heavily in algorithmic precision to keep users engaged, user satisfaction depends much more on immediate contextual relevance and a felt sense of personal agency than on historical accuracy alone.
Context
Algorithmic personalization has become the defining feature of modern digital platforms like TikTok, Spotify, and Netflix. While these recommendation systems are critical drivers of user engagement and commercial success, academic research has historically focused on the technical backend precision of algorithms rather than how users subjectively experience them. This creates a gap between an algorithm's technical performance and a user's subjective evaluation, as highly precise systems can still cause user frustration if they fail to align with psychological expectations, immediate contextual needs, or the user's desire for agency.
Objectives
This thesis addresses this by investigating which perceived dimensions of algorithmic personalization predict overall user satisfaction. Grounded in Expectation Confirmation Theory, the central research question asks how the perceived factors of relevance, accuracy, and user control in algorithmic personalization influence a user's overall satisfaction with a digital platform. Additionally, the study explores which individual dimension holds the strongest predictive power and evaluates whether satisfaction levels differ across short-video, music streaming, and long-format video platform environments.
Methodology
The study applied a quantitative, cross-sectional survey design using validated Likert-scale items. Data were collected via Qualtrics, and after data quality filtering and an attention check, a final analytical sample was established. The sample was distributed across short-video users, music streaming, and long-format video. The data were analyzed using multiple linear regression in R, assessing the independent dimensions of Perceived Control, Perceived Accuracy, and Perceived Relevance while controlling for platform type.
Findings
The overall regression model proved highly significant, indicating that the set of predictors collectively explains a meaningful portion of the variance in user satisfaction. Perceived Relevance emerged as the strongest and most significant predictor of user satisfaction, followed closely by Perceived Control. Crucially, Perceived Accuracy did not reach statistical significance, indicating that accuracy carries negligible independent weight once contextual fit and user control are factored in. Furthermore, platform type showed no significant effect on satisfaction.
For platform operators, the strategic takeaway is clear: to maximize user satisfaction, product investments should prioritize user-facing control mechanisms and real-time contextual responsiveness over a narrow focus on backend historical accuracy.