Why E-Learning Fails the 70-20-10 Test

The Ogre Always Wins

In the middle of a Code Combat level, a student stopped typing. He didn’t ask me for the syntax of a for loop. He asked, “Why does the ogre keep winning even when my code is right?”

I opened my mouth to explain variable scope — the formal answer, the lecture answer — and caught myself. He didn’t need a lesson. He needed to fight the ogre again, differently, and lose in a new way until the pattern clicked. He needed the 70%.

In the mid-1990s, Michael Lombardo and Robert Eichinger — building on research they’d conducted with Morgan McCall at the Center for Creative Leadership in the 1980s — put a number on something most of us sense but rarely act on. They surveyed roughly two hundred successful executives about where their capabilities actually came from. The answer was a ratio: 70% of meaningful learning comes from challenging on-the-job experiences. 20% comes from relationships, mentoring, and feedback. And just 10% comes from formal training — the courses, the workshops, the carefully produced slide decks.

The 70-20-10 model was built for corporate leadership development, not classrooms. But the more I taught — a user interface design course, a game-based coding curriculum, a film studies class — the more I recognized its shape everywhere. The assignments where students came alive weren’t the ones where I lectured well. They were the ones where I gave them a problem just beyond their reach and stood back.

What troubled me was the inverse recognition. When I looked at e-learning — the industry I was building my career inside — I saw an entire ecosystem built almost exclusively around the 10%. Video lectures. Interactive slides. Multiple-choice quizzes. A three-hundred-billion-dollar industry flowing into the thinnest slice of the model while the other ninety percent went undesigned.

This is the story of what happened when I tried to flip that ratio — in three very different courses, with three very different results. One succeeded almost by accident. One required constant pushing. And one failed in ways I’m still trying to understand.

A Model That’s Precisely Wrong and Directionally Right

How did you learn to do what you do?

In the mid-1990s, researchers at the Center for Creative Leadership put that question to two hundred successful executives. The answers coalesced into a ratio that would reshape corporate training for decades. Seventy percent of their development, the executives reported, came from challenging on-the-job assignments. Twenty percent came from relationships — mentors, peers, the boss who told them the truth they didn’t want to hear. Ten percent came from formal training: the workshops, the courses, the classrooms.

It’s a clean story. Maybe too clean.

Here’s what the ratio’s champions rarely mention: those numbers came from retrospective self-reports. Successful people looked backward over their careers and guessed at what made them successful. The sample was small, homogeneous, and pre-selected for the outcome. When researchers Kajewski and Madsen went looking for the empirical foundations in 2012, they concluded bluntly: “There is a lack of empirical data supporting 70:20:10.” The phrase itself never appears in the original Lessons of Experience (1988) — it was crystallized later, a convenient shorthand that hardened into doctrine.

Other researchers found ratios closer to 55-25-20 — the numbers shift with industry, role, and what you’re trying to learn. Real learning, it turns out, resists clean fractions.

So should we throw the model out?

No. Because the direction holds up remarkably well.

Strip away the specific percentages and you find a hierarchy that aligns with decades of independent learning science. Kolb showed that concrete experience drives deeper understanding than abstract instruction alone. Bandura demonstrated we acquire complex behaviors through observation, modeling, and feedback from others. Lave and Wenger revealed that learning is fundamentally situated — it happens in the doing, among people who are doing it together.

The 70-20-10 model didn’t discover these principles. It stumbled into them sideways, through anecdote rather than experiment. But it pointed in the right direction: experience first, relationships second, formal instruction third.

Think of it as a compass, not a GPS. A compass won’t give you turn-by-turn directions, and if you try to navigate by precise bearing in a dense forest, you’ll walk into trees. But it will keep you oriented. It will stop you from walking south when you need to go north.

For an industry that has been walking south for two decades, even a rough compass would be revolutionary.

The Greenhouse Problem: Why E-Learning Gets It Backwards

Here is the dirty secret of the e-learning industry: we have spent two decades perfecting the art of building greenhouses and calling them education.

Walk into any Learning Management System and you will find the same thing. Slides with tasteful gradients. Videos where a disembodied voice narrates bullet points over stock footage of people shaking hands. Knowledge checks that ask you to recall what you just read thirty seconds ago. Everything is climate-controlled. The temperature is perfect. The lighting is soft. And the roots of every plant inside are rotting from neglect.

The structural bias is baked into the medium itself. LMS platforms are used as content delivery systems masquerading as learning environments. Their entire architecture — upload a video, attach a quiz, track a completion — is optimized for the 10%. For formal instruction. The thing that research tells us contributes the least to how adults actually learn. We have built a three-hundred-billion-dollar industry around the smallest slice of the pie and then acted confused when the completion rates hover somewhere between “abysmal” and “statistical noise.”

And those completion rates — the 5% to 15% that the industry treats as normal — are not the disease. They are the fever. The body rejecting something it knows is wrong. Learners are not lazy. They are not distracted. Many never intended to finish — they were browsing, sampling. But those who did intend to finish and still dropped out? They were bored. They could feel, in their bones, that watching a forty-minute video about “conflict resolution strategies” was not going to teach them how to handle the conversation they were dreading on Monday morning. So they clicked away. They closed the tab. They did what any rational person does when handed a recipe for bread but told they cannot touch the flour: they went to find a kitchen somewhere else.

I know because I built some of those greenhouses myself.

The reason this keeps happening is not stupidity. It is economics. Producing content is easy. It scales. You can record one video and push it to ten thousand employees by Thursday. Designing a meaningful challenge — one that is appropriately difficult, personally relevant, and scaffolded so that failure teaches rather than discourages — is brutally hard. It requires knowing your learners. It requires iteration. It requires accepting that what you build might not work the first time.

So we default to the greenhouse. We water the plants on schedule. We keep the pests out. And we wonder why, when our learners finally step outside into the wind and the rain of their actual jobs, they have no idea how to stand on their own.

The greenhouse is not useless. Seeds need to germinate somewhere. But a greenhouse that never opens its doors is not growing anything. It is preserving specimens.

Into the Forest: Designing the 70%

The forest does not care about your lesson plan.

Wind does not wait until you have finished the module on “aerodynamic stress.” Pests do not consult the syllabus before arriving. In a forest, a tree either grows roots deep enough to hold, or it falls. There is no partial credit. And this is precisely why the trees that survive are stronger than anything a greenhouse could ever produce.

Designing the 70% — the experiential core — means accepting this. It means building courses where the challenge is the curriculum, not a decorative capstone bolted onto the end. It means designing for the scar.

I learned this by accident with Code Combat, an after-school coding camp for middle schoolers. The entire architecture of the course is challenge-first: students write real code to navigate real levels, and the levels do not care about their feelings. An ogre does not politely wait while you debug your for-loop. A maze does not simplify itself because you looked frustrated. The learning is in the struggle — the moment when a student stares at a screen full of red error messages and has to decide whether to quit or to think harder.

Not every student found the sweet spot. Some hit the wall and went silent — those silences were where I failed to scaffold fast enough.

What Code Combat taught me is that the 70% is not “practice.” It is not the exercise at the end of the chapter. It is the chapter itself. The challenge is the content. And the design challenge for the instructor is not “what information do I deliver?” but “what problem do I set loose, and in what order?”

The ordering matters enormously. Too easy and the student coasts. Too hard and they shatter. The art is in the gradient — each level slightly beyond current ability, each failure producing just enough frustration to fuel the next attempt. Game designers have understood this for decades. Csikszentmihalyi called the sweet spot “flow” — where challenge is just high enough to stretch skill without overwhelming it. Hamari and colleagues (2016) found that perceived challenge was a particularly strong predictor of learning outcomes — stronger than immersion, and mediated by engagement. The sweet spot is not comfort. It is productive discomfort.

Code Combat taught me the principle by accident. The UX design course — a graduate-level class at SJSU — was where I tested it on purpose. Instead of front-loading theory and saving the “real” project for the final weeks, I made every single assignment a mobilization of the student’s full abilities. Week two was not “read about typography.” Week two was “design a poster for a real event with real constraints, and your classmates will critique it on Thursday.” Every week was the forest. Every week had wind.

The difference was immediate. Students stopped treating assignments as hoops and started treating them as problems they owned. They argued about typeface choices in the hallway. They redesigned things that had already been graded, not for extra credit but because the first version bothered them. Something like the IKEA effect — the tendency to overvalue what we build with our own hands — in full bloom. Even the wobbly parts.

This is what “designing for the scar” means in practice. You do not protect learners from difficulty. You engineer difficulty — carefully, deliberately, with enough scaffolding that the fall teaches instead of destroys, but never so much scaffolding that the fall becomes impossible. The forest is not cruel. It is honest. And honesty, it turns out, is what makes roots grow deep.

The Mycelium Layer: Engineering the 20%

Beneath the floor of an old-growth forest runs an invisible architecture. Mycorrhizal networks — fungal threads thinner than a human hair — connect tree to tree, ferrying nutrients and chemical alarm signals across vast distances. Ecologist Suzanne Simard’s research showed that these networks don’t just exist; they redistribute resources, channeling nutrients from strong trees to struggling ones. A forest without its mycelium is a collection of isolated organisms. A forest with it is a community.

The 20% works the same way. Social learning — mentoring, peer feedback, collaborative struggle, the casual conversation that unlocks a stuck problem — is the connective tissue that turns isolated learners into a learning ecosystem. And like mycelium, it’s invisible when it’s working and conspicuously absent when it’s not.

Here’s the uncomfortable truth about the 20%: it doesn’t happen unless you push for it. I learned this the hard way across three courses. In my Code Combat course, mentoring became foundational — but only because the structure demanded it, not because students spontaneously sought each other out. In my UIX design class, forming teams proved transformative for learning outcomes, and I kicked myself for not doing it sooner. Peer testing — students evaluating each other’s interfaces — broke the ice in ways no icebreaker exercise ever could. The discussion forums generated genuine social learning, but only when there was something real to discuss, not when participation was simply required.

The question I kept circling was: What digital tools actually make this scale?

The research confirms what I saw in my courses: social tools work when they’re woven into real tasks, and they become ghost towns when they’re not. I watched it happen — the forums where students debriefed genuine challenges hummed with activity, while the ones with manufactured prompts collected tumbleweeds. Peer mentoring boosts motivation — though the confident students benefit first, which means designers have to be intentional about routing nutrients to the struggling trees themselves. The mycelium feeds the strong first; you have to intervene to make sure the rest get connected.

Discussion forums work when they’re tied to authentic tasks — debriefing a challenge, sharing a solution, critiquing a peer’s approach. Wikis work when there’s a shared artifact to build. Virtual meetings work when they’re small enough for everyone to speak. The pattern isn’t about which tool; it’s about whether the tool creates genuine interdependence or just proximity.

This is where most e-learning platforms fail the 20% entirely. They bolt on a discussion board the way a builder might scatter mushrooms on a concrete floor and call it an ecosystem. But mycelium needs the right conditions: moisture, organic matter, root systems to connect to. Digital social learning needs the equivalent — shared challenges that create genuine need for each other, structures that make reaching out easier than going it alone, and enough psychological safety to admit when you’re the struggling tree that needs nutrients sent your way.

The 20% is the hardest component to engineer and the easiest to neglect. You can’t lecture it into existence. You can’t automate it. You can only build the conditions and then — critically — push. Form the teams earlier than feels comfortable. Make peer testing a requirement, not an option. Create discussion prompts that demand sharing real work, not performing engagement. The mycelium will grow — but only if the conditions are right. No one gets to skip the hard part of building them.

The Amplifier and the Admission

Here is the part that complicates the narrative: the 10% matters more than its size suggests.

In my Code Combat course, students would grind through level after level, building loops and conditionals through sheer repetition — the forest doing its work. But periodically, they’d hit a wall. Not a difficulty wall — a conceptual wall. They could see the ogre. They could see their code. They could not see why one refused to fall to the other.

That’s when I’d lecture. Ten minutes, maybe fifteen. Variable scope. Function returns. The kind of formal instruction that, on paper, contributes just 10% to learning. But in practice, those ten minutes didn’t teach a lesson — they unlocked the previous three hours of struggle. Students would return to the same level with the same code and suddenly see what they’d been looking at all along.

Researchers at the Center for Creative Leadership describe this as an “amplifier effect.” Formal training is small in volume but disproportionate in impact — not as a foundation, but as a catalyst. It clarifies what experience has already half-taught. Without the struggle, the lecture is abstract. Without the lecture, the struggle is noise. The 10% is not the least important slice. It is the lens that brings the other 90% into focus.

Which brings me to the admission.

I taught film intro — an elective at De Anza College, a lecture course in a large auditorium. The catalog defined it before I arrived: lecture, occasional Q&A, midterm, final. I understood the 70-20-10 model by then. I believed in it. And I failed to make it work.

The UX design course had clear challenges — design a prototype, test it with real users, iterate under pressure. Code Combat had its levels. But film? I couldn’t find the equivalent. Watching a film is not doing it the way writing broken code is doing programming. Analyzing cinematography is closer to the 10% than the 70%, no matter how sophisticated the analysis. I could lecture beautifully about Hitchcock’s use of sound, and my students would nod and forget.

The 70% for film was always there — it was production, not analysis. Picking up a camera, shooting a scene badly, watching it back, shooting it again. I designed a course around watching and discussing films, not making them. I think I know what went wrong — I never built the forest. But I’m not fully certain that’s the whole story.

Part of it is structural. I had inherited the format, not chosen it. A two-hundred-seat auditorium does not lend itself to peer critique sessions or iterative studio work. The UX course ran with a small cohort in a discipline where critique is already the culture. Code Combat was an after-school program — informal enough that the format was whatever the challenge demanded. Very few instructors get that kind of latitude. Most teach inside a room that already has a name, a seat count, and a catalog entry. The 70-20-10 model, applied honestly, is rarely about designing the ideal course from scratch. It is about finding the degrees of freedom inside the course you actually have — and pushing as hard as those constraints will allow.

The 70-20-10 model didn’t fail me in film intro class. I failed it. The point was always the question behind the ratio: Have you designed a challenge worth struggling through? If the answer is no, no amount of social scaffolding or polished lectures will save you. And if the answer is yes — as it was in Code Combat, as it grew to be in UX design course — the model almost takes care of itself.

The compass works. But you still have to walk into the forest.