The 70-20-10 Learning Model: Does It Still Hold Up for L&D?

Where the Numbers Actually Come From

The 70-20-10 model is most associated with Morgan McCall, Robert Eichinger, and Michael Lombardo’s research at the Center for Creative Leadership (CCL) in Greensboro, North Carolina. The seminal work appears in Lombardo and Eichinger’s 1996 book The Career Architect Development Planner, which drew on studies CCL had been conducting since the late 1980s.

The methodology matters for understanding what the numbers mean. CCL researchers surveyed high-performing managers and executives and asked them to reflect on their most important developmental experiences. Respondents described, in retrospect, the relative contributions of different learning types. The 70-20-10 proportions emerged as rough central tendencies from those self-reports.

This is survey data about memory and attribution, gathered from a specific population (successful managers), in a specific era (pre-internet, pre-LMS), about a specific skill type (leadership and managerial effectiveness). It was never a controlled experiment. It has never been replicated with a representative sample of workers across functions and skill types. The researchers themselves — Lombardo and Eichinger in subsequent interviews — have noted that the numbers were meant as a “rule of thumb” and have expressed frustration at how literally the model has been applied.

None of this makes the model useless. It means the model should be evaluated accurately.

What the Research Actually Supports

Strip away the precise percentages, and the 70-20-10 model makes a claim that learning research broadly supports: most meaningful professional learning happens through doing, in context, with feedback from others — not primarily through structured training events.

This is consistent with decades of research on expertise development. K. Anders Ericsson’s work on deliberate practice demonstrates that expertise in complex domains accumulates through effortful, feedback-rich practice in the domain itself — not primarily through passive instruction. Etienne Wenger’s communities of practice research shows that social learning embedded in professional work is a primary mechanism for knowledge transfer and norm acquisition. Neither of these bodies of research endorses the specific 70-20-10 ratios, but both support the model’s directional claim that formal training is a relatively small part of how professionals actually develop.

The practical implication that many L&D professionals draw from this — and which is defensible — is that training programs designed as discrete events, disconnected from application opportunities and social reinforcement, will underperform. The model provides a useful corrective to organizations that conflate “training completions” with “development.”

Where the Model Misleads

The problems arise when the ratios are treated as prescriptive targets rather than descriptive approximations.

Skill type matters enormously. The 70-20-10 proportions were derived from leadership development data. For technical skills, compliance training, medical procedures, safety protocols, or other domains where getting it wrong is costly, the appropriate ratio of formal instruction is almost certainly higher. A nurse learning medication dosing calculation, an engineer learning a new structural analysis method, or a finance professional learning a new regulatory framework cannot rely on “learning by doing” in a 70% proportion without unacceptable error rates. Applying 70-20-10 uniformly across skill types leads to dangerous underinvestment in formal instruction for high-stakes technical domains.

The “70%” is not self-executing. Experiential learning does not happen automatically just because someone is on the job. It requires structured reflection, psychological safety to admit and analyze mistakes, and feedback mechanisms. Organizations that interpret “70% experiential” as permission to reduce formal learning investment without increasing the quality of on-the-job learning structures get worse outcomes, not better ones. The model implies a rich developmental environment at work; many workplaces don’t have one.

The 20% is often the hardest to create deliberately. Social and peer learning is valuable, but “put people in proximity and learning will happen” is not a design. Communities of practice, mentorship programs, and social learning networks require intentional design and maintenance — often more effort than formal training programs. Treating the 20% as automatic because it’s “informal” misses the investment required to make it effective.

Practical Applications for L&D Professionals

Used as a design philosophy rather than a measurement formula, 70-20-10 offers L&D professionals three actionable orientations:

• Design for transfer, not just training events. If formal training accounts for only a fraction of development, the job of L&D is to design for what happens after the course — the stretch assignments, manager conversations, and peer networks that turn training into capability.
• Invest in the “20%” deliberately. Social learning doesn’t happen by accident in most organizations. Cohort-based learning, structured peer consultation, and community platforms require active design and facilitation. Organizations that want the learning return from the 20% need to invest in it.
• Match learning modality to skill type. Use 70-20-10 thinking to push back against over-reliance on formal training for leadership and soft-skill development. Use it with caution for technical or safety-critical skills where formal instruction has a higher appropriate weight.

This aligns with how modern corporate training strategy frameworks increasingly approach capability building — moving away from event-based training toward integrated learning ecosystems. The 70-20-10 model, for all its empirical limitations, helped push L&D discourse in that direction, and that contribution is real even if the numbers are not.

Having the Honest Conversation

The best L&D practitioners I’ve encountered are clear-eyed about the model’s limitations and use it accordingly. They cite it as “a useful framing, not a measurement target” when presenting to senior leaders who may have seen it cited as fact. They push back when budget decisions are made by applying the ratios mechanically to time or resource allocation. And they acknowledge that the most important question isn’t whether the numbers are right — it’s whether their organization is investing in all three learning types intentionally and well.

That kind of honest professional judgment is harder to put on a slide than a pie chart showing 70-20-10. But it’s the judgment that actually leads to better learning investment decisions.