By Andreas Pettersson, Founder, Leaders ADAPT
Every leadership development plan I have been handed lately has the same three numbers on it. Seventy percent on the job, twenty percent from other people, ten percent in a classroom. The 70-20-10 rule is so familiar that nobody asks where the numbers came from, and the answer is more interesting than the numbers.
I hired more than 80 people while scaling a company to 150, and I have watched the ratio work and fail. It works when a leader treats the 70 as a delegation problem. It fails when learning and development treats it as a budget split. This page covers the origin, what the evidence shows, and how I run the 70 with a manager who needs to grow.
Quick answer: The 70-20-10 rule is a leadership development guideline from the Center for Creative Leadership holding that roughly 70 percent of learning comes from challenging on-the-job experience, 20 percent from developmental relationships such as managers, mentors and coaches, and 10 percent from formal coursework. The ratio comes from a 1996 survey of executives describing how they believed they learned, so it is a guideline rather than a measurement.
What is the 70-20-10 rule in leadership development?
The 70-20-10 rule is a leadership development guideline holding that about 70 percent of a leader's learning comes from challenging on-the-job experience, 20 percent from developmental relationships, and 10 percent from formal coursework and training. It is used to plan how a person's growth time gets spent, not to grade where it went.
One disambiguation, because the numbers collide. This is not the 70 percent decision rule, which says you should make most calls once you have about 70 percent of the information you would like. That one is a decision-speed rule and I cover it on the CEO decision system and the 70 percent rule. This page is about how people learn to lead.
Who created the 70-20-10 model?
Three researchers connected to the Center for Creative Leadership: Morgan McCall, Michael Lombardo and Robert Eichinger. The ratio was published in The Career Architect Development Planner (1996) and grew out of CCL's Lessons of Experience research, which asked executives how they had learned to lead. CCL still owns the framing and calls it on its own site "a research-based, time-tested, classic guideline," and later on the same page a rule.
Guideline is accurate. Rule is marketing.
Is the 70-20-10 rule backed by research?
Partly, and less than the precision of the numbers suggests. Here is what the record shows.
- The ratio comes from a 1996 survey of approximately 200 executives who were asked to reflect on how they believed they had learned, according to the model's own documented history (Wikipedia, "70/20/10 model (learning and development)," accessed 2026). That is self-report from a small, senior sample, not a measurement of learning outcomes.
- CCL's current page on the model gives no sample size, method or outcome data, and describes the ratio as emerging from "over 30 years" of research rather than from a specific study (ccl.org, "The 70-20-10 Rule for Leadership Development," accessed 2026).
- The learning scholar Andy Jefferson's assessment, as recorded in the same documented history, is that the model "is neither a scientific fact nor a recipe for how best to develop people."
- A 2021 debate article in Public Money & Management (Taylor & Francis) questions the rule's evidence base directly, and is worth reading before you buy training on the strength of the ratio.
My reading, as someone who has built the development plans and not just read about them: the ratio is a memory aid. The direction is right. People learn to lead mostly by leading, somewhat from the people around them, and a little from being taught. Anyone who says it is exactly 70 is quoting a 1996 survey with a confidence the survey never had.
What the criticisms of the 70-20-10 model get right
Three criticisms hold up. The numbers were never measured, so treating them as targets is false precision. The model says nothing about quality; a bad on-the-job year counts as 70 percent too. And it is silent on the one thing that decides whether the 70 works, which is whether anyone above the learner knows how to delegate real work and let it stay delegated.
What counts as the 70 percent in the 70-20-10 model?
| Share | What the model says | What it usually means in practice | What it means when it works |
|---|---|---|---|
| 70 percent, experience | Stretch assignments and challenging roles, on-the-job learning, informal learning | The person keeps their normal job and gets a project on top | Real outcome ownership, with authority, and a cap on how often they can hand it back |
| 20 percent, relationships | Managers, mentors, coaches, peers, feedback | An annual review and an occasional mentor coffee | Structured feedback on the reasoning behind decisions, twice a week, plus proximity to someone better |
| 10 percent, formal | Courses, workshops, reading | The thing that gets budget because it has an invoice | Base information, delivered before the stretch, and not mistaken for the development itself |
The middle column is why the model has a reputation problem. The last column is how I run it.
How do you apply the 70-20-10 rule to develop a manager?
Run the 70 as delegation, not as exposure. The sequence I use:
Delegate the outcome, not the task. State the outcome you want, the feeling you want around it, and the deadline, and let the person choose how to get there. Give them 70 percent of the clarity and let them fill in the last 30 through their own problem-solving; the full method is on my delegation training page. A leader who specifies the how has delegated typing, not judgment, and judgment is the only thing the 70 percent was ever supposed to build.
Set an interim checkpoint when you hand it over. Every missed handoff is upstream: your delegation was not good enough. Name the checkpoint at the start, a Wednesday status check for a Friday deadline, so the person can win early instead of confessing late.
Cap the escalations. For someone newly handling decisions with real stakes, I give neither open-ended discretion nor none. I give a capped number of escalations out of their full load, on the order of one in nine, so they have to build judgment about what genuinely needs me.
Then I pair the cap with a structured debrief, about twice a week and usually inside the one-on-one meeting, where the question is why a case was handled the way it was, not whether the outcome was right. The debrief is the 20 percent. Without it the cap is just a rule.
Build a sandbox for the weaker area. Where someone is strong at one part of a process and weak at another, I define a bounded project for the weak part and delegate the details inside it, rather than expecting them to design the project. Starters build from nothing and system thinkers refine what exists; very few people are both.
Measure iterations, not years. I have met people who did the same job for ten years and have one year of experience, and people who did ten chaotic things in five years and are decades ahead. The unit of development is real cycles, so structure the year for cycles: shorter projects, faster decisions, more handoffs.
Expect the disappointment. The curse of the intelligent delegator is that your brain assumes people can do what you can do, and you will be disappointed when they cannot. The answer is not lowering the bar. It is growing the people, which is the whole point of the ratio.
For the 20, one rule from my own scaling years: put a hungry new hire next to an already strong performer rather than adding headcount in isolation. Both raise their effort because neither wants to be the weaker one, and the strong one becomes the developmental relationship without a mentoring program ever being announced. I would not have sustained my own company's growth without coaches and mentors, and I still keep a leadership psychologist and sit in a peer group, because you cannot always be the teacher.
For the 10, my position is unfashionable. Training, training, training, and then nothing happens, because a book, a course or a class gives you base information and never gets you anywhere until you act. I once built a leadership course of 300 slides and nobody cared; what people used were five short practices they could run on Monday.
Set the expectation before any formal training: awareness today, practice this month, habit in about three months. Anyone who tells you a single workshop changes everything is lying to you.
Does the 70-20-10 model still work with AI and remote work?
The direction holds and the boundary between the buckets moves. AI has compressed the 10: base information that used to take a course now takes an afternoon, and a company can turn one veteran's judgment into a reusable skill new hires inherit. AI has not touched the 70. Nobody learns to hold a hard conversation or let a delegated project stay delegated by asking a model.
Remote work strains the 20 hardest. In an office the developmental relationship happened by proximity; remote, it has to be scheduled, which is why the twice weekly debrief above is a calendar entry and not an intention. The types of coaching that fill the 20, and the 7 leadership traits the 70 is meant to produce, have their own pages.
The 70-20-10 rule FAQ
What is the 70-20-10 rule in leadership development?
The 70-20-10 rule is a leadership development guideline holding that about 70 percent of a leader's learning comes from challenging on-the-job experience, 20 percent from developmental relationships such as managers, mentors and coaches, and 10 percent from formal coursework and training. Organizations use it to plan how development time and budget are allocated. It is a guideline for emphasis rather than a measured breakdown of how learning occurs.
Who created the 70-20-10 model?
The model is attributed to Morgan McCall, Michael Lombardo and Robert Eichinger, researchers connected to the Center for Creative Leadership. It was published in The Career Architect Development Planner in 1996 and grew out of CCL's Lessons of Experience research, which asked executives to reflect on how they had learned to lead. CCL continues to publish and promote the framework as part of its leadership development work.
Is the 70-20-10 rule backed by research?
The ratio originates in a 1996 survey of about 200 executives who self-reported how they believed they learned, not in a measured study of learning outcomes. CCL describes it as a research-based guideline but publishes no sample, method or outcome data on its page. Critics, including the learning scholar Andy Jefferson, describe it as neither a scientific fact nor a recipe. The direction is widely accepted; the exact percentages are not evidence-based.
What counts as the 70 percent in the 70-20-10 model?
The 70 percent covers learning from doing: stretch assignments, new responsibilities, leading a project, taking over a team, handling a crisis or a turnaround. For it to develop judgment rather than workload, the assignment needs real outcome ownership, the authority to make decisions inside it, a defined checkpoint before the deadline, and a limit on how often the work can be escalated back to a senior person.
What is the difference between the 70-20-10 rule and the 70 percent decision rule?
They share a number and nothing else. The 70-20-10 rule describes how leaders learn: 70 percent from experience, 20 from relationships, 10 from training. The 70 percent decision rule describes when leaders should decide: once about 70 percent of the desirable information is in hand, because waiting for more usually costs more than an occasional wrong call. One is a development guideline; the other is a decision-speed heuristic.
Does the 70-20-10 model still work with AI and remote work?
The direction still holds. AI has compressed the formal 10 percent, since base information that once required a course can now be acquired quickly, and it has not replaced experiential learning, which still requires making real decisions with real consequences. Remote work makes the 20 percent harder because developmental relationships no longer form by proximity, so feedback and debriefs have to be scheduled deliberately.
Keep the direction, drop the decimals
Use the 70-20-10 rule for what it is good at: reminding you that the course is the smallest part and the job is the biggest. Then stop quoting the numbers and check the thing the model never measured: whether the person you are developing owns an outcome, has a checkpoint, and gets asked why. Do that with one manager this quarter and you will have run the 70-20-10 rule better than most companies that put it in a slide.
The system the 70 runs on
Every step above is a delegation habit, and most leaders lose delegation habits within a month because the work drifts back onto their desk. The 5-Minute Leader is the system I wrote to stop that drift: a delegation protocol with a three-minute handover conversation, a one-on-one protocol that makes the debrief a fixed part of every meeting, an accountability protocol for the moment someone hands the work back, and one short test inside the delegation protocol that decides whether a task should leave your desk at all. Run it on one person first. The ratio takes care of itself.
Andreas Pettersson hired more than 80 people while scaling an AI company past 150 as one of Canon's youngest CEOs, cutting his own week from 60 hours to 45 in the process. He founded Leaders ADAPT to help CEOs delegate without losing control.
