Learning Analytics
Learning Analytics
What is learning analytics?
- Analyzing learning data to assess and improve training
- Looks at completion, engagement, achievement, and behavior
- Manage training with data, not gut feeling
What learning analytics really changes is the conversation in the quarterly training review. Instead of 'This course was well received,' people say 'Dropouts clustered in week 3, and that is the only week with a 47-minute video.' Once evidence is on the table, debate turns into an improvement task. It also changes how the training team defends its budget — with logs, not excuses.
Analysis is usually split into four layers. Descriptive analytics shows what happened (completion rate, progress, login time); diagnostic analytics digs into why (drop-off points, wrong-answer rate per question); predictive analytics flags who is at risk (the likelihood that someone inactive for 2 weeks won’t finish); and prescriptive analytics decides what to do (personalized content recommendations). Most companies in Korea stop at the first layer. They pull a completion-rate table every month but often never open up the segments where dropout clusters.
It also helps to be clear about the boundaries. Learning analytics works with data from the learning process. People analytics, which also covers hiring, turnover, and promotion, differs in both data source and purpose. For a course with fewer than 30 learners, interviews are more accurate than statistics. The smaller the sample, the more one or two dropouts swing the percentages, so for small courses, use conversation instead of numbers.
The most common misconception is that installing a dashboard starts the analysis. A tool provides screens; people decide what to look at. More dangerous is reading correlation as causation. High performers may simply have taken more training, rather than training having made them high performers. Make completion rate an individual KPI and more people will play a video while doing other work; let word spread that data is a surveillance tool and people start avoiding the learning logs themselves.
Anna ChoExpert in organizational culture, coaching, emotion management, and CS consulting
John KimInstructor John Kim, CEO of Nanum Consulting and author of B2B Marketing DesignAs e-learning and LMS adoption built up a store of learning logs, it took shape as a field in the 2010s. It developed where educational technology meets data analysis.
Consider a hypothetical case: Company A, a retailer with 2,400 employees and KRW 410 billion (≈ US$306,748,500) in annual revenue. It had run an online program for newly appointed team leaders for three years, but the completion rate would not budge from 41%. All anyone said was, 'The content must not be good.'
- Kim, a manager on the HR development team, downloaded about 12,000 learning logs from the past two years out of the LMS and split them by week. 38% of all dropouts were concentrated in week 3 alone.
- The playback data showed that learners stopped, on average, at the 11-minute mark of the 47-minute week-3 video. A 12-minute case video sat in exactly that spot.
- When the team called five managers who had dropped out, three gave the same answer: "I watch it on the subway commute, and I can’t keep going past 15 minutes."
- Park, a team lead in training planning, split a 47-minute video into six 8-minute clips and added a hands-on worksheet. Learners who have been inactive for two weeks now get reminders sent in the name of their division head.
- Six months later, the completion rate rose from 41% to 78%, and week-3 dropout fell from 38% to 9%. Only after linking to HR data did they first see that performance-review conversations were held 2.3 times as often for completers as for non-completers.