Reading group insights: What works in online learning? Lessons from 100 courses

colourful speech bubbles

For our July 2025 session of the Digital Education Reading Group, we discussed a recent study that examined the relationship between online course design and student performance in a community college in the United States. Leonard Houx introduced the paper, Unlocking Success: Key Features of College Online Pedagogical Practices that Predict Better Performance, which investigates which elements of course design correlate with student achievement. This is an increasingly relevant topic as the University of London moves toward more flexible models like micro-credentials.

A rare look at real-world online teaching data

The study analysed 100 fully online courses from a single institution in the southeastern US, using a dataset of over 3,600 student enrolments. It considered both student and instructor variables, offering an unusually detailed look at what correlates with success in online learning. The researchers focused on three design dimensions: scaffolding, student agency, and presence/interactivity.

Leonard highlighted how these constructs draw from long-standing educational theories, from Vygotsky to Garrison, though he also noted the absence of frameworks from cognitive science, such as Mayer’s multimedia principles.

“Light-touch instructor presence, clear structure, and opportunities for informal peer interaction emerged as surprisingly powerful design features.”

What the data revealed: headlines and surprises

Among the design features that correlated most strongly with better outcomes were:

  • Clearly stated learning objectives
  • Varied media formats (e.g. combining video, text, and audio)
  • Regular announcements and reminders
  • Moderate—not heavy—tutor presence
  • Informal opportunities for student-to-student interaction

Some of these findings confirmed our expectations. Others, like the weak correlation between tutor feedback and outcomes, challenged assumptions and sparked thoughtful debate.

Context matters: who are the learners?

Much of the discussion centred on the specific learner context: US community college students, many of whom are studying part-time, returning to education, or have lower prior attainment. Several participants cautioned against generalising the findings to all higher education contexts, especially where students might seek more academic challenge and intellectual engagement. While the study focused on preventing failure and encouraging persistence, it told us little about what helps high-achieving students excel.

Designing for diverse learners: flexibility vs structure

The group reflected on the balance between giving learners agency and offering clear structure. There were differing views on how much choice to give students, particularly when they may not yet have developed strong self-regulated learning strategies. We discussed ideas from cognitive load theory, emphasising the importance of tailoring support for novice versus experienced learners; another reminder that one-size-fits-all design is rarely effective.

Practical implications and institutional resonance

The findings resonated with insights from our own student experience surveys. An analysis of over 5,000 survey responses showed that students value clear structure, accessible information, and consistent support: hallmarks of good scaffolding.

We discussed the potential for low-effort, high-impact interventions, such as automated weekly announcements or AI-assisted feedback summaries which are scalable whilst preserving a sense of tutor presence.

Final reflections

Leonard concluded by noting that while the study may not provide definitive answers, it does offer “real-world validity”, a rare and valuable lens on what actually happens in large-scale online teaching. As we continue to design for increasingly diverse learners, these kinds of insights help ground our work in both evidence and empathy.

The findings reflect a specific learner context—open-entry, low-attainment community college students, so applicability to other higher education settings requires careful consideration.

Sincere thanks to Leonard for leading and presenting this session, and to all who joined the conversation.

This summary was co-written with generative AI, drawing directly from the session transcript.