Scholarlink

<p>Digital marketing analytics capture behavioral outcomes—clicks, dwell time, conversions—but offer limited insight into how engagement unfolds during interaction. This research develops a multimodal framework that infers engagement-related behavioral configurations from synchronized gesture, gaze, facial, vocal, and linguistic signals, across two overlapping samples of video-mediated sessions (<i>N</i> = 34 and <i>N</i> = 40). Stress-related activation is robustly and negatively associated with attentional stabilization, a result that withstands collinearity diagnostics, bootstrap resampling, and alternative model specifications. Its positive association with intentional engagement is also observed at the session level, but is substantially weaker once features shared across indicators are removed and does not hold at finer temporal resolution—underscoring that not all engagement associations are equally robust to measurement overlap. Multimodal indicators add explanatory power beyond attention-based metrics within the framework, though this should be read as internal incremental contribution pending external validation. This research offers a transparent behavioral inference framework and a methodological template for testing the robustness and temporal structure of such indicators, and identifies external validation as the next step toward establishing them as measures of engagement.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Measuring digital engagement dynamics through multimodal behavioral inference: a marketing analytics framework

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年7月8日

查看全文
  • Charlotte De Sainte Maresville,
  • Christine Petr,
  • Olfa Haggui,
  • Felipe Restrepo

摘要

Digital marketing analytics capture behavioral outcomes—clicks, dwell time, conversions—but offer limited insight into how engagement unfolds during interaction. This research develops a multimodal framework that infers engagement-related behavioral configurations from synchronized gesture, gaze, facial, vocal, and linguistic signals, across two overlapping samples of video-mediated sessions (N = 34 and N = 40). Stress-related activation is robustly and negatively associated with attentional stabilization, a result that withstands collinearity diagnostics, bootstrap resampling, and alternative model specifications. Its positive association with intentional engagement is also observed at the session level, but is substantially weaker once features shared across indicators are removed and does not hold at finer temporal resolution—underscoring that not all engagement associations are equally robust to measurement overlap. Multimodal indicators add explanatory power beyond attention-based metrics within the framework, though this should be read as internal incremental contribution pending external validation. This research offers a transparent behavioral inference framework and a methodological template for testing the robustness and temporal structure of such indicators, and identifies external validation as the next step toward establishing them as measures of engagement.