Every learning and development team has a slide with three numbers on it: 70:20:10. We learn 70% by doing the job, 20% from other people, and 10% from formal training.
It’s the most widely adopted planning model in L&D, and the most specific, which is unusual in a field that primarily deals in principles.
The appeal is obvious. A profession that rarely quantifies anything was handed three numbers that add up to a hundred and tell you exactly how to split a budget.
In 2024-2025, US corporations and educational institutions with more than 100 employees spent $102.8 billion on training. A model that tells you where to funnel those funds is worth a great deal. Of course, this makes it worth asking whether the numbers are real.
This article covers where they came from, what happened when someone finally tested the framework in practice, and what the evidence says about each of its three parts. It’s time for a closer look.
What Is the 70:20:10 Model, and What Does It Claim?
Stated plainly, the model makes three claims about where professional capability comes from.
- 70% From Experience: Doing the job. Stretch assignments, unfamiliar problems, new responsibilities, and the mistakes that come from all three. This is the leg most people mean when they cite the model.
- 20% From Other People: Coaching, feedback, observation, working alongside someone better, and the conversation after a meeting goes badly. In practice, this is the leg most often reduced to a mentoring scheme.
- 10% From Formal Learning: Courses, eLearning, qualifications, reading. The part L&D has traditionally owned, and the part the model implies is worth a tenth of the effort.
So far, so reasonable. The trouble starts when you ask what the percentages are percentages of. After all, learning is not a quantity with an agreed unit. Is the claim:
- That 70% of learning events are experiential?
- That 70% of the time someone spends learning happens on the job?
- That 70% of the capability they end up with was built through experience?
These are three distinct claims. Each would need a different study to establish. The model does not say which one it is making, and neither does anybody citing it. Of course, this would be a technicality if the model were only ever used to describe, but it’s not. It is used to allocate.
The ratio gets used to justify spending on the things that support experience: coaching, performance support, platforms, reflection. That is a sensible use of money. But it’s not what the numbers say. It’s just what practitioners decided the numbers ought to mean.
What Does the 70:20:10 Model Get Right?
Before taking the numbers apart, it’s worth being clear about why this model remains popular, because the answer is not that the L&D profession is gullible.
Throughout the 1980s and the 1990s, corporate learning was almost entirely course delivery. Anyone who suspected that people mostly learned their work by doing their work had no way to make that case in a planning meeting. The 70:20:10 framework gave them one. Three things follow:
- It named a real problem. Budgets sat almost entirely in formal delivery, and nothing in the profession’s vocabulary challenged that.
- It gave L&D a number. Principles do not survive a finance director. Ratios do.
- It points somewhere the evidence supports. The direction holds up even where the proportions do not.
The achievement was political rather than scientific, but it was still an achievement.
On that third point, the strongest support comes from the institution the original research came from. The Center for Creative Leadership later interviewed 393 executives across China, Singapore, India, and the United States about what had developed them.
Formal courses and training “were mentioned by just 9%-15% of interviewees… and were not among the top seven sources of leadership learning.” That’s a much larger sample, and formal training still lands roughly where the model puts it.
Its most prominent advocate has never claimed more than that. Charles Jennings: “It’s important to be aware that 70:20:10 is a reference model and not a recipe. The numbers are not a rigid formula.”
Read that way, the model is a diagnostic rather than a target. Audit where your effort and budget actually go. If 90% of both sits in formal delivery, you’ve found a mismatch worth exploring and explaining.
Who Created the 70:20:10 Model?
The ratio first appears not in a journal but in a practitioner’s tool. The Career Architect Development Planner, published in 1996, describes itself as offering “95 research based and experience tested development plans and coaching tips”.
Behind it sat earlier interview work from the Center for Creative Leadership: The Lessons of Experience, published in 1988 by Morgan McCall, Michael Lombardo, and Ann Morrison.
In this work, successful executives were asked to look back and name the experiences that shaped them. One version of the question ran:
“Please identify at least three key events in your career, things that made a difference in the way you manage now.”
Hold that question in mind, because everything else follows from it.
That research produced a taxonomy of developmental experiences. It did not produce a ratio. The ratio is a later summary of it, written for people planning careers rather than for people testing a hypothesis.
Why It Felt Established Anyway
A rule of thumb with no study behind it should not survive thirty years of scrutiny. However, this one did, because it kept appearing to be confirmed.
Alan Clardy went looking for that confirmation and found five separate research traditions that all seemed to land on the same number:
“Five different literature traditions examined here seemingly converge on the claim that 70% or more of work-based learning occurs informally… Yet a detailed examination of this literature finds that the evidential basis for the 70% rule is weak.”
Five bodies of work arriving from different directions at roughly the same figure sounds like real corroboration. It’s also what coincidence looks like when several fields measure loosely defined things in incompatible ways.
That’s a polite way of putting it. Clardy, on the other hand, puts it down to: “sloppy scholarship, inconsistent conceptualizations, and fundamental research protocol problems”.
Why That Method Cannot Produce a Ratio
The original survey question was a good one. It’s just not the kind of question that generates proportions.
- It Measures Memorability, Not Learning: Ask people what shaped them and you get what they can recall: the vivid, the difficult, the emotionally loaded. A stretch assignment that nearly ended a career is unforgettable. A compliance course does not share that quality. The method is biased towards drama, and drama happens on the job.
- It Samples Only The Survivors: The numbers come from a survey of successful executives. Nobody asked the people who had the same assignments, the same bosses, the same courses, and stalled anyway. Without that comparison, you cannot tell whether challenging experience cause the success or simply accompanied it.
- It Counts Nominations, Not Contribution: Ask for three key events, and someone might name two assignments and a boss. That is a tally of nominations. Turning it into “70% of learning” means assuming each nomination represents an equal quantity of learning, which nobody has established.
Then there’s a fourth problem, and it’s the one that should worry L&D most. Formal training that works stops feeling like training. Once a model becomes the way you think, you no longer experience it as something you were taught. In other words, the “10%” may simply be measuring how invisible good formal learning becomes over time.
None of that makes the ratio malicious. It makes it a reasonable summary of a set of interviews, promoted into a planning rule it was never built to be.
Does the 70:20:10 Model Actually Work?
For a framework adopted this widely, one thing is remarkable. Until 2018, nobody had checked whether it worked.
Then Johnson, Blackman and Buick did. They interviewed 145 people across four Australian government bodies. The question was whether implementing 70:20:10 actually moved learning into managerial capability. It did not.
“Results showed that despite middle managers’ awareness of, and willingness to take part in, ongoing skill development, attempts to develop capability through learning transfer by implementing the 70:20:10 framework were not achieving the desired outcomes.”
Why is this? The authors identify four misconceptions. Each one is a design error you can go looking for in your own programmes.
- Unstructured Experience: Handing someone a stretch assignment is not development. Unstructured experiential learning does not automatically build capability. The “70” was being treated as something that simply happens rather than something you design for.
- Narrow Social Learning: Managers interpreted social learning as formal arrangements, usually mentoring, which missed the everyday modelling that does most of the work.
- Unsupported Formal Training: Courses ran, then people returned to workplaces with no support for applying any of it. The “10” was delivered and then abandoned.
- Separate Systems: Experience, social, and formal were treated as three streams to be resourced rather than one system, so nothing reinforced anything else.
Their conclusion was as follows: social learning should act as the “glue” connecting the other two. Not a fifth of the budget. The connective tissue.
Of course, this is qualitative research in one country’s public sector, so it tells us how the framework fails rather than how often. The authors are equally direct about the wider problem, describing the framework’s “atheoretical nature and the lack of empirical evidence to support its effectiveness.”
What to Use Instead of the 70:20:10 Ratio
Each leg of the ratio points at something real. All three have been measured properly, and the measurements give better instructions than the purported proportions ever did.
Experience: Exposure is Not Development
The “70” is usually read as good news, because experience is something employees are already having. But, as we’ve seen, unstructured experience does not automatically build capability. What decides it is the conditions around the work.
Blume, Ford, Baldwin, and Huang pooled 89 samples covering roughly 12,500 people to find out what actually predicts formal training reaching the job.
| What Predicts Transfer | Correlation |
|---|---|
| Cognitive ability | .37 |
| Voluntary participation | .34 |
| Supervisor support | .31 |
| Conscientiousness | .28 |
| Transfer climate | .27 |
| Post-training knowledge | .24 |
| Motivation | .23 |
There is no clear lever, only a set of small additive ones. Experiential learning alone is not a silver bullet, not with all these variables to consider. Indeed, the strongest situational factor, voluntary participation, is a decision about how you enrol people rather than how you teach them.
The authors state plainly that there is “no clear superiority of individual variables over situational variables, or the reverse,” and warn that work-environment correlations halve, from .54 to .23, once you stop letting the same person rate both the cause and the effect.
In other words, while “70” may be the largest number in the model, it rests on the thinnest evidence of the three.
Other People: The Strongest Workplace Evidence You Have
Taylor, Russ-Eft, and Chan pooled 117 studies of behaviour modelling training, which is the “20” turned into an actual method.
Effects on what people knew and could do were around a full standard deviation, putting the average trained person ahead of roughly 84% of the untrained. Effects on what they did back at work were around a quarter of that.
This gap is important. To close it learners need to see bad models as well as good ones, practise on scenarios they generate themselves, set clear goals, train alongside their managers, and be rewarded for their behaviour.
None of that survives a culture where admitting ignorance is risky. Frazier and colleagues pooled 136 samples covering more than 22,000 employees and found psychological safety tracking learning behaviour at .62. This is among the strongest relationships in organisational research.
That is what Johnson’s “glue” looks like with numbers attached.
Formal Learning: Transfer Does Not Fail, It Decays
The “10” has justified neglecting formal training for thirty years. It’s usually supported by a second piece of folklore: that only 10% of training transfers to the job.
That figure traces to a single 1982 article where it appears as an off-the-cuff estimate voiced by a hypothetical training director. There’s no study, no method, no data.
The real picture is more useful. Saks and Belcourt asked professionals at 150 organisations how much of their training was still being applied. The results were:
- 62% immediately afterwards
- 44% at six months
- 34% after a year
Formal training does not fail at the classroom door. It decays, on a timescale you can schedule against. That’s a design problem, not a justification for pulling your budget.
How to Apply 70:20:10 in Practice
Having spent an article taking the numbers apart, keep in mind that what follows are guidelines, rather than a strict formula. That said, they are what the evidence supports. The research (as detailed above) tells you which conditions matter and which direction to move in.
Start with the audit. Take last quarter’s L&D effort and sort it into the three buckets. Not based on budget, but based on effort. If almost all of it sits in formal delivery, you have found which way your gap runs. You can then build your to-do list accordingly.
- Design, Don’t Just Assign: Before a stretch assignment starts, agree what capability it should build and how anyone will know whether it did. Book the debrief in the same calendar invite as the assignment itself. Without a defined outcome and a scheduled review, you’ve given someone workload, not a development opportunity.
- Go Beyond Mentoring: Formal mentoring reaches a few people slowly. The everyday modelling that does most of the work happens in questions, comments, and worked examples from peers. So give people somewhere to ask, somewhere to show their work, and visible credit when they help someone else.
- Build The Landing Strip: Train the manager alongside the learner and brief them on what they will be asked to support. Blume’s two most actionable predictors, supervisor support and transfer climate, are both decided before anyone opens a module.
- Connect The Three: Formal learning should end in something the learner applies, the application should get discussed with peers, and that discussion should shape what gets taught next. Run as three separate streams and each one is likely to underperform.
That last point is as much a platform problem as a design one. Content in one system, conversation in another, and reinforcement in a spreadsheet is exactly how the leak happens.
Keeping the formal learning, the social layer, and spaced reinforcement in one place is the whole design intent behind the Impact Suite.
Final Words
70:20:10 is not a finding. It’s a rule of thumb from a 1996 development planner, built on interviews that asked successful people what they remembered. It was then promoted into a budget rule by thirty years of repetition. Clardy’s verdict, after examining the literature, was that it “needs to be set aside.”
Thankfully, we can set the ratio aside and keep the instruction that lies underneath it. Formal training is a small part of how people become capable. Experience and other people do most of the work. That much the evidence supports. It also tells us which way the gap runs, which is all we were ever really using it for.
What replaces the numbers is not just another set of numbers. It’s a shorter list of things that demonstrably move performance: enrol people who chose to be there, train the managers who have to support them, and build a climate that rewards the new behaviour.
None of that needs a ratio. It just needs someone who’s willing to stop taking the “70” for granted.
The Impact Suite is built to deliver experiential, social, and formal learning at scale. Book a demo to see it in action, or download our “Social Learning Guidebook” to unleash a bounty of actionable insights and practical tips.
