The impact of short videos on student performance in an online-flipped college engineering course

The study investigates whether short pre-recorded instructional videos improve student engagement and academic performance in a fully remote, online-flipped college engineering course during the COVID-19 pandemic. With the rapid shift to remote modalities, video became central to learning, yet quantitative evidence on optimal video length in credit-bearing, flipped engineering contexts is scarce. Drawing on theories such as cognitive load and multimedia learning, and prior findings from MOOCs and K-12 settings that suggest shorter videos increase engagement, the authors test if similar benefits extend to a credit-bearing, online-flipped engineering drawing course. The purpose is to quantify effects on engagement (video viewing time) and learning outcomes (final exam scores) and to gather student perceptions, thereby informing course and video design in remote flipped engineering education.

The literature background contrasts flipped versus asynchronous online teaching, emphasizing flipped learning’s active, student-centered activities and potential for enhanced motivation, engagement, and knowledge retention. Prior work identifies multiple video characteristics affecting effectiveness (e.g., length, presentation style), with video length being a particularly tractable factor. Studies (e.g., Guo et al., 2014) show short videos (6–10 minutes) increase viewing time in MOOCs, and K-12 research notes higher engagement with shorter videos though not always improved test performance. Other studies (Afify, 2020; Manasrah et al., 2021) suggest short videos can improve achievement and are more engaging. Flipped models may improve long-term knowledge retention via active learning and interactions. Engagement and motivation are critical to learning outcomes, with flipped classes often outperforming traditional methods. However, most prior work focuses on MOOCs, language learning, or on-campus settings rather than fully remote, credit-bearing engineering courses—motivating the present study’s context-specific investigation.

Design: A fully remote, online-flipped Introduction to Engineering Drawing course at Suzhou City University was implemented. Weekly long lecture videos (~55 min, matching regular class length) were recorded via the platform’s recording tool. Short videos were created by trimming the same content into segments of 7–9 minutes (average 8 minutes), each covering one complete knowledge point. All materials were uploaded to a virtual learning platform; students had one week to study pre-class material, followed by a 2-hour synchronous interactive session (discussions, Q&A, tutoring). Homework was collected via the platform. Participants: 65 first-year mechanical engineering students (Jiangsu Province admittees with similar Gaokao profiles) participated; repeaters were excluded. The registry randomly assigned 35 to the long-video group and 30 to the short-video group. Average Gaokao scores were comparable (317.9 ± 7.6 vs. 319.3 ± 7.8), indicating similar intrinsic learning skills. Measures: The platform automatically recorded per-student video viewing behaviors (clicks and total viewing time). Engagement was quantified as video viewing time, normalized by total assigned video length (values >1 indicate rewatching). One proctored final exam (two hours, simultaneous for both groups, camera invigilation) provided quantitative performance data. Questionnaires: A 9-item Likert survey (5-point scale: strongly disagree to strongly agree) included 3 pre-course items (internet skills and attitudes) and 6 post-course items (viewing behavior and perceptions of video length). Reliability was high (Cronbach’s alpha = 0.87). Statistical analysis: Due to non-normal distributions, Mann-Whitney U-tests compared engagement times and exam scores between groups; significance set at 0.05. One-way ANOVA compared Likert responses. MATLAB was used for analysis. Implementation details: Short videos were edited with Adobe Premiere Pro (720p, H.264, mp4 audio). Although the system could not verify continuous on-screen attention, students were instructed not to play videos unattended. Groups were randomized to mitigate bias.

• Engagement: Short videos significantly increased engagement. Median normalized engagement time was 0.932 (short) vs. 0.747 (long), a 24.7% improvement. Short-video group showed lower variance and more students watching ≥75% of content. Mean normalized engagement for the short-video group exceeded 1 (indicating rewatching), while the long-video group was <1.
• Exam performance: The short-video group achieved higher final exam scores. Median and average scores were 7.4% and 9.0% higher, respectively, than the long-video group. The difference in average scores was statistically significant (p = 0.0334 < 0.05), with the short-video group displaying smaller standard deviation (more consistent performance).
• Pre-course attitudes and skills: No significant differences between groups in internet skills (Q1) or initial attitudes toward the subject and format (Q2–Q3), all p > 0.05, suggesting comparable baseline readiness and motivation.
• Viewing behaviors and perceptions: Long-video students reported slightly higher frequency of pausing/rewinding (Q6: 4.58 vs. 4.13, p > 0.05) and stronger agreement that videos were too long (Q7: 4.06 vs. 3.13, p > 0.05). Perception that videos were too short was similar between groups (Q8: 3.16 vs. 2.89, p > 0.05). Despite similar self-reported ability to stay focused (Q4: 3.86 vs. 4.06, p > 0.05) and retain information (Q5: 4.06 vs. 4.10, p > 0.05), objective exam scores favored the short-video group.
• Contextual comparison: Engagement times in this credit-bearing, flipped course exceeded typical MOOC engagement durations reported in prior work, likely due to degree stakes and structured flipped activities.

The findings support the hypothesis that shorter videos enhance both engagement and achievement in a credit-bearing, fully remote flipped engineering course. Increased engagement in the short-video group aligns with cognitive load theory: segmenting content into self-contained, concise units reduces extraneous load, enabling better processing and schema construction. The statistically significant exam score advantage indicates that improved engagement translated into measurable learning gains. Pre-course equivalence in internet skills and attitudes suggests observed differences are attributable to video length rather than baseline motivation or readiness. Compared with prior MOOC-focused research, higher engagement in this setting underscores the influence of course credit, assessment stakes, and active, student-centered flipped sessions. Measuring engagement via objective viewing time enhances validity over solely self-reported engagement metrics. Overall, short videos, when integrated into a well-structured flipped format, can positively affect learning behaviors and outcomes in engineering education.

In a fully remote, credit-bearing flipped engineering drawing course, short instructional videos markedly improved student engagement (median +24.7%) and enhanced academic performance (short group’s median and average exam scores were 7.4% and 9.0% higher than the long-video group). These results are consistent with cognitive load theory and indicate that concise, knowledge-point-focused videos help students process information more effectively. Engagement levels exceeded those typically seen in MOOCs, suggesting that credit-bearing contexts and active flipped sessions further support sustained viewing. The study offers actionable recommendations for designing short, coherent video segments linked to in-class activities and incorporating pauses for processing. Future research will broaden behavioral metrics, explore additional video-design factors (e.g., embedded quizzes, multimedia), examine learner contexts and devices, and replicate across more courses to strengthen external validity.

• Attention verification: The platform could not confirm continuous, attentive viewing; students might have played videos without watching despite instructions.
• Single-course, single-institution context: Findings stem from one engineering course at one university during pandemic lockdown, limiting generalizability.
• Sample size and randomization constraints: Although groups were randomly assigned and baseline Gaokao scores were similar, residual individual differences and unmeasured confounders may remain.
• Self-report measures: Some perceptions and behaviors were captured via self-report Likert items, which may be subject to bias.
• Modality and content specificity: Results may depend on engineering content that lends itself to segmentation into discrete knowledge points; effects may vary across disciplines and course designs.