My Take on the 70:20:10 Model: Why Ratios Don’t Make Great Learning Strategies

Dec 1

Introduction

The 70:20:10 model occupies a strange place in the learning and development world. It is everywhere - quoted in conference keynotes, embedded in leadership programs, invoked by consultants, and repeated by managers who want to emphasize that “real learning happens on the job.”

If you’ve worked in L&D for any length of time, you’ve likely heard the familiar formula:
“People learn 70% of what they know from experience, 20% from others, and only 10% from formal training.”

The model is presented as if it reflects some objective truth about how humans learn at work. It sounds credible. It aligns with what many practitioners observe. And it’s simple, seductively so, with its three-part structure and round numbers that tickle something in our pattern-loving brains.

But once you peel back the surface, the numbers turn out to be surprisingly shaky. They came from retrospective interviews with executives, were later interpreted into a ratio, and eventually became a kind of industry folklore. The model’s power has less to do with research evidence or large-scale measurement of workplace learning and more to do with how neatly it reinforces a shift that was already happening in workplace learning.

This article doesn’t intend to “debunk” 70:20:10 for the sake of it. Rather, I want to contextualize the model: where it came from, why it stuck, what the research actually says, and why learning professionals should be wary of using ratios as strategic guideposts. More importantly, I want to share my own perspective as a learning designer who sees enormous value in practice-based learning - but also understands why the percentages themselves rarely help us make better decisions.

What I want to convey is that overreliance on the model, especially its numerical breakdown, can distract practitioners from the deeper question:

What kind of learning design actually works?

What the 70:20:10 Model Claims

In its common form, the 70:20:10 model states that workplace learning is distributed roughly as follows:

70% - experiential learning: on-the-job tasks, challenges, real problems, trial-and-error
20% - social learning: coaching, mentoring, feedback, teamwork, observation, guidance from peers and managers
10% - formal learning: structured instruction such as courses, workshops, or e-learning

These categories map loosely onto how we might observe adult learning in daily work. Employees do not spend most of their time in classrooms or LMS modules. They spend it doing tasks, interacting with colleagues, and figuring out problems as they arise. Formal training occupies the smallest slice of the workweek.

Jefferson and Pollock from the Association for Talent Development note in their blog post titled “70:20:10: Where Is the Evidence?” that the model’s popularity is rooted in this intuitive alignment with everyday experience. It feels true.

The intuitive appeal is precisely why 70:20:10 spread so widely. It reinforces a message many L&D practitioners have long championed:

Learning is not an isolated event; it is a continuous, embedded process.

But intuitive models are not always accurate ones. And the neat division into three categories can obscure the complexity of real learning ecosystems.

The Strange - and Vague - Origins of the Model

When you trace the model’s history, you quickly discover that the numbers emerged through interpretation rather than research.

Retrospective executive interviews

The ratio traces back to the Center for Creative Leadership’s Lessons of Experience (McCall, Lombardo & Morrison, 1988). The researchers asked executives to describe formative career experiences - not to quantify learning, but to reflect on what shaped their leadership.

But the famous ratio does not appear anywhere in the book.

Years later, Lombardo and Eichinger summarized these narratives in The Career Architect Planner (1996), stating that lessons “tend to come” from tough jobs (70%), other people (20%), and courses or reading (10%). It was meant as a rough heuristic - a simplified interpretation, not as a universal principle.

Alan Tough and the “about 70%” misconception

Educational psychologist Alan Tough suggested that “about 70%” of adult learning occurs outside institutions, but this was not workplace-specific and was never meant to be part of a ratio.

No empirical origin found

Kajewski & Masden (2012) explicitly investigated the model’s roots and concluded there was:

no clear origin, and
no empirical data supporting the exact percentages

In other words, the numbers didn’t emerge from measurement - they emerged from storytelling.

And yet, the model continues to shape corporate learning, not because of evidence but because it reinforces a broader shift in how we think about learning.

Why the Idea Still Matters: Learning Isn’t an Event

Even without strong empirical grounding, the 70:20:10 model resonated - and continues to resonate - because it captured a growing realization: learning does not equal instruction.

Several major educational and psychological movements were already pushing learning in this direction. The model simply gave the trend a catchy numerical frame.

The shift away from traditional instruction

For decades, workplace training followed a fairly predictable pattern:

Deliver content (lecture, reading, presentation)
Test retention
Assume transfer will happen naturally

This approach often led to information overload, lack of engagement or motivation, low retention, and minimal behavioural change. The disconnect between “knowing” and “doing” became increasingly visible, especially in complex or fast-changing roles. Practitioners and researchers alike began questioning whether traditional training formats could meaningfully influence performance.

Constructivism: Learning through experience

Constructivist theory - especially influential from the 1980s onwards - argued that adults learn by:

making decisions
solving problems
applying knowledge in real contexts
reflecting on outcomes
integrating new information into existing mental frameworks

This philosophy underpins many modern L&D practices: scenario-based learning, case studies, simulations, project-based tasks, and reflective practice.

In this sense, the “70%” of the model echoes a long-standing belief: experience is a powerful teacher, especially when it’s structured, supported, and evaluated.

Cognitivism and Bloom’s Revised Taxonomy

From a cognitivist perspective, we learn by actively processing information, drawing on our memories, and using what we already know. Bloom’s taxonomy and its revision by Lorin Anderson (a former student of Bloom’s) address the cognitive domain of learning.

The framework highlights the progression from remembering and understanding to applying, analyzing, evaluating, and creating. These higher-order skills cannot be developed through passive instruction alone.

When you map Bloom’s higher levels to the model, you see emphasis similar to that of the 70:20:10:

Experiential learning maps to “apply” and “analyze”
Social learning contributes to “evaluate”
Formal learning is strongest at “remember” and “understand”

The model didn’t create this hierarchy - it merely echoed it in a simplified form.

Why the model became so appealing

By the early 2000s, several forces converged:

more dynamic workplaces
increasing need for continuous learning
the rise of coaching and mentoring practices
recognition that performance was shaped by experience, environment, and culture, not just instruction
a growing appreciation of informal learning

The 70:20:10 model captured these ideas in a single, memorable frame. It didn’t need scientific precision to become influential; it simply aligned with a widespread shift already in motion.

And that is precisely why the model is still used and still matters: it reminds us that learning is lived, social, contextual, and continuous, not confined to courses.

But, as we’ll see, the model offers very little guidance on how to support that kind of learning.

**What Evidence Does Exist?**

Despite its popularity, rigorous research on the model is nearly nonexistent. To date, only one scientific study has examined it directly: by Johnson, Blackman & Buick (2018), Human Resource Development Quarterly.

Will Thalheimer of Work-Learning Research reviewed the study, highlighting a few notable points:

Conducted within the Australian public sector
Qualitative interviews with senior and middle managers
Participants had been encouraged to adopt 70:20:10 in middle-management development
Focused on implementation, not validating the ratio

Key finding: Model did not deliver expected learning transfer

Four issues emerged:

Overreliance on unstructured experiential learning
Narrow, inconsistent interpretations of social learning
Assumptions that behaviour would change after training without follow-up
Lack of integration between the three components

Where implementation faltered

Experiential learning was often unstructured and unsupported
Social learning depended on individual preference - rarely designed systematically
Formal training was disconnected from workplace application
Managers lacked time, guidance, or frameworks to support learning

The research did not validate the model; rather, it highlighted how difficult it is to implement successfully.

Overall, the study suggests a simple truth: the model fails not because the idea is wrong, but because organizations tend to reduce it to numbers instead of designing the infrastructure needed to support real experiential and social learning.

Other Ratios: 55-25-20 and 45+45+10

If 70:20:10 reflected an underlying truth about how adults learn, one might expect new research to reinforce it. Instead, alternative models propose drastically different proportions.

Training Industry’s 55-25-20

Their 2018 research suggested that workplace learning distribution might look more like:

55% experiential
25% social
20% formal

This already challenges the idea that formal learning is only “10%.”

Swinburne Edge’s 45+45+10 (Eva Kyndt, 2023)

Based on 15+ years of research, Kyndt proposed a very different distribution:

45% experiential
45% social
10% formal

Its core insights:

Social learning is more significant than traditionally recognized.
Learning must be intentional and connected - ad-hoc informal experiences are ineffective.
Training requires pre-work, follow-up, and workplace integration.

The continued emergence of alternative models highlights the main point:

The numbers are not fixed truths; they vary by industry, job type, culture, and organizational context.

The Problem with Ratio Thinking

So, why does ratio-based thinking persist if the numbers themselves are unreliable? Because ratios feel concrete. They appear to offer guidance. But in practice, they can be counterproductive.

1. Ratios oversimplify learning ecosystems

They imply that learning can be neatly measured and categorized when, in reality, experiential, social, and formal learning constantly overlap.

2. They distract from what makes learning effective

Organizations often believe they should force learning into proportions, fearing that too much formal training violates the “rule.” But that’s the thing about good learning design: it relies on relevance, timing, scaffolding, feedback, and application, not adherence to a percentage.

3. They encourage reactive rather than strategic design

It becomes easy to justify weak experiential learning with “that’s the 70%” - even when no structures for support or reflection are in place.

4. They’re especially outdated in a digital-first environment

Modern learning is deeply intertwined with technology. This aligns with connectivist perspectives, where knowledge and learning emerge through networks: digital tools, platforms, communities, and environments. Today, a single workflow might combine experiential, social, and formal learning seamlessly.

In other words, learning is less of a pie chart and more of a constellation.

Which brings us to the next point.

AI, Knowledge Networks, and the Future of Workplace Learning

The 70:20:10 model predates remote work, digital ecosystems, and AI, all of which fundamentally reshape how learning happens.

Employees now learn through:

real-time performance support
contextual information embedded in tools
AI-driven suggestions and nudges
knowledge-sharing platforms
chatbots and virtual assistants
remote collaboration tools
microlearning at the point of need

This is experiential + social + formal, all blended in one cohesive, continuous learning flow.

In an AI-enhanced environment, the boundaries between the “70,” “20,” and “10” blur even further. A single modern blended workflow may incorporate:

experiential learning (solving a real issue)
social learning (collaborating or asking for feedback)
formal learning (referencing a micro-module or a corporate wiki article)

The boundaries between categories collapse. If anything, technology makes rigid ratios even less meaningful.

The future of learning is integrated, not segmented.

My Take: What Actually Works

After reviewing the research, observing learner behaviour, and designing multiple types of workplace education, my conclusion is this:

The most useful element of the 70:20:10 model is not the ratio, but rather, it’s the reminder that learning is multifaceted and context-dependent.

What actually drives performance?

1. Integration

Formal + Social + Experiential learning should form a cycle, not separate boxes.

2. Contextualization

Learners need relevance, practice, and feedback aligned with real problems.

3. Support systems

Managers, mentors, peers, culture, and digital tools all influence whether learning sticks.

4. Timing

Training has the greatest impact when delivered at the moment learners need it.

5. Design quality

Scenario-based tasks, problem-solving activities, coaching structures, and application opportunities matter more than any ratio.

6. Avoiding arbitrary constraints

Trying to engineer learning experiences to “fit” a percentage distracts from real business and learner needs. In short:

Ratios do not design learning. People do.

As a learning experience designer, I see far more value in designing for authenticity and relevance than in categorizing activities according to a model created decades ago from retrospective interviews.

Final Recommendations

If you choose to reference the 70:20:10 model, treat it as an idea, not an equation.

To build effective learning ecosystems:

Focus on the real needs of learners, not proportions
Design experiences where formal, social, and experiential elements reinforce each other
Provide structure for informal learning (coaching, reflection, communities of practice)
Support managers so they can support learning
Enable performance with digital tools and real-time resources
Use theory, evidence, and organizational context to guide decisions
Treat learning as a continuous process, not a set of isolated events

Modern learning is too complex, too dynamic, and too intertwined with technology to be represented by a single ratio.

The model is a starting point for conversation - not a roadmap.

A thoughtful and context-sensitive approach will always serve learners - and organizations - better.

References

concepts and theoriesIMHOpractical LXD

Yuliya Romanova