I Learned Two Cloud Topics Two Ways. One-Month Retention Test.

"Hands-on labs beat passive video" is repeated constantly in cloud training advice, including by me, in other posts, without ever being shown. So I ran a small test on myself: two cloud topics I had never touched before, roughly matched for difficulty, the same total study time on each, one learned entirely through video content and notes, the other entirely through hands-on labs. Then I quizzed myself at one week and again at one month, honestly, before looking anything up.

Caveat up front: this is a sample size of one, run on myself, with all the bias that implies. It is not a controlled study. Treat it as one data point, not proof, and notice whether the pattern matches your own experience rather than taking the numbers as universal.

The setup

# Two topics, matched for approximate difficulty, 6 hours of study time each
topics = {
    "Azure Virtual Networking":   {"method": "video_course", "hours": 6},
    "Azure Storage & Access Control": {"method": "hands_on_labs", "hours": 6},
}

# Self-quiz: 20 practical questions per topic, scored honestly, no lookups
# Scored at two checkpoints: 1 week after study, 1 month after study
results = {
    "Azure Virtual Networking":      {"week_1": 15, "month_1": 8},   # out of 20
    "Azure Storage & Access Control": {"week_1": 17, "month_1": 15}, # out of 20
}

for topic, scores in results.items():
    decay = scores["week_1"] - scores["month_1"]
    decay_pct = decay / scores["week_1"] * 100
    print(f"{topic}: {scores['week_1']}/20 -> {scores['month_1']}/20  ({decay_pct:.0f}% decay)")

Azure Virtual Networking: 15/20 -> 8/20  (47% decay)
Azure Storage & Access Control: 17/20 -> 15/20  (12% decay)

What actually happened

The video-learned topic started slightly weaker at the one-week mark, which already hints at something, and then fell off a cliff by one month: nearly half of what I thought I knew at week one was gone a month later. The lab-learned topic started stronger and barely moved. By the one-month mark the gap was not subtle: 8 out of 20 versus 15 out of 20 on topics I had rated myself equally confident in at the one-week checkpoint.

Why this probably isn't a fluke, even at n=1

The honest mechanism, as far as I can tell from paying attention to my own recall during the quiz, was not that the video content was lower quality. It was that I could recall facts about networking concepts without being able to reconstruct how to do anything with them, while the lab-learned storage and access control material came back as procedure, not facts. Procedural memory built from actually typing commands and watching them succeed or fail seems to survive a month in a way that passively-absorbed facts do not. That matches a decent amount of general learning-science literature on the difference between declarative and procedural memory, for whatever that is worth here.

What I'd actually change based on this

Not "never watch another video," that would overcorrect past what one data point supports. What I would change: use video for the initial orientation to a new concept, the first pass that gets the vocabulary and the shape of the problem into your head, then move to hands-on labs as fast as possible for anything you actually need to retain past a week. The video-only topic above was pure video with no lab component at all, which is probably the worst version of the comparison, and a more realistic recommendation is video-to-orient, labs-to-retain, not one or the other exclusively.

If you want to see what this looks like applied to a real exam rather than an isolated topic, the complete lab-first study plan I built for the AZ-104, six weeks of nothing but Azure CLI commands mapped to the actual exam domains, is the practical version of the argument this post is making in miniature.

---

I put together the complete case for why hands-on training beats theory-first content for cloud specifically, with the research and the practical format that makes it work, over here: Hands-On Labs vs Theory: Practical Cloud Training Approach.

Has your own retention matched this pattern, or have you had a topic that stuck fine from video alone? Drop it in the comments. 👇