📍 Independent. Unsponsored. Reliable.

The 70-20-10 Model for Learning: Does It Still Hold Up in 2026?

The 70-20-10 model for learning holds up as an idea and fails as a formula. As of November 2026, the claim that most workplace development happens through experience and other people is plausible. But no …

Dark graphic titled The 70-20-10 Model for Learning with a balance scale, a ladder and a clipboard checklist.

The 70-20-10 model for learning holds up as an idea and fails as a formula. As of November 2026, the claim that most workplace development happens through experience and other people is plausible. But no source reviewed for this guide shows that the split is 70, 20 and 10 percent. The model’s own champions say it is not a recipe.

That matters if someone has asked you to “apply 70-20-10” to a budget, a strategy deck or a competency framework. The numbers can start a useful conversation about where people really learn. They cannot tell you how many dollars, hours or courses to fund.

This guide traces where the model came from and weighs the evidence for and against it. Then it gets practical. You get a table of what the numbers can and cannot justify, a diagnostic table for auditing your own learning mix, and a rule for when to drop the percentages and keep the idea.

Where did the 70-20-10 model come from?

The model is usually traced to Center for Creative Leadership (CCL) research on how executives develop, published in the 1988 book Lessons of Experience. The 70-20-10 format itself came later, in a 1996 planning tool by Lombardo and Eichinger. The phrase does not appear in the book.

CCL describes the book as based on “information from 191 accomplished and high-potential executives,” covering key developmental events and the lessons drawn from them, on its Lessons of Experience product page. An ATD article from July 2014 adds that “the phrase never actually appears in the book,” and that Lombardo and Eichinger summarized the findings in the Career Architect Planner (1996). Some popular summaries date the model to the 1980s, so expect sources to disagree on the year.

What question did the executives actually answer?

According to the same ATD article, the executives were asked: “Please identify at least three key events in your career, things that made a difference in the way you manage now.” That is a request to recall turning points. It is not a measurement of how learning time was divided, and it had no comparison group of less successful managers.

Our reading, not a claim from those sources: a tally of remembered events cannot tell you where learning time or learning effect sits. It also says nothing about frontline staff, new hires or regulated roles, because the sample was senior executives.

What does the 70-20-10 model for learning actually say?

It says people develop mainly through work experience (70), through working with and learning from others (20), and through formal courses and programs (10). The people who promote it describe the numbers as a reminder of where learning happens, and say the ratio will vary with the work and the results needed.

The three channels are usually described like this:

  • 70, experience: learning through the work itself, including practice, experimentation and reflection.
  • 20, social: learning through coaching, mentoring, networks and collaboration.
  • 10, formal: learning through planned courses and programs.

The 70:20:10 Institute’s own model page says “the specific ratio (70:20:10), in any given situation, will vary.”

A post on Charles Jennings’ blog (the URL carries a June 2013 date) is blunter. It says “70:20:10 is a reference model and not a recipe” and “the numbers are not a rigid formula.”

Notice what that does to the question you were asked. If the model’s own advocates say the ratio is not fixed, a brief that says “make it 70-20-10” asks you to apply something they say should not be applied that way.

Is there evidence for the 70-20-10 model?

Not for the percentages. The sources reviewed describe the model as retrospective and built on executives’ self-reports. A 2012 paper, as quoted by ATD, found a “lack of empirical data supporting 70:20:10,” and a 2018 critical review concluded the 70% rule “needs to be set aside.” The direction of the idea fares better than the numbers.

The table below separates what each source supports from what it does not.

Source What it is What it supports What it does not support
CCL, Lessons of Experience (1988) Study of 191 accomplished and high-potential executives Executives name on-the-job events as formative Any measured split of learning time or effect
ATD article (2014) Critical overview, quoting Kajewski and Madsen (2012) Formal coursework “was sometimes included by executives” as a difference-making event The claim that 70-20-10 is fact: it calls it neither “a scientific fact nor a recipe”
Clardy (2018) Review of five literatures behind the 70% rule of informal learning Informal learning matters at work The weight of the 70% figure: it finds “the evidential basis for the 70% rule is weak”
70:20:10 Institute and Jennings Advocates explaining intended use Using the model as a prompt to look beyond courses Using the numbers as a rigid formula

Two limits on this table. I read the Kajewski and Madsen finding only as quoted by ATD, not in the original paper. And Clardy’s abstract, which I read through a library catalog record, covers the wider “70% rule” across five literature traditions, not only the CCL work. I also found no newer study in my search that validates the split, but that does not prove none exists.

Can you use the 70-20-10 model to set a learning budget?

No. The numbers describe where people say they learn, not how to spend. The Institute’s model page, as read for this guide, presents the percentages as sources of learning and does not mention budgets. A 70% line for “experience” has no obvious product to buy, and a 10% line for formal learning could starve programs that need depth.

Use the table to decide which requests the numbers can and cannot support.

Request you may receive Can 70-20-10 justify it? Why Use instead
Split the budget 70/20/10 No The numbers describe remembered sources of development, not cost or return Fund by performance gap and cost per outcome
Spend 70% of learning time on the job No Most working time is already on the job; the issue is whether that time builds skill Check whether the job gives practice, feedback and support
Cut formal training to 10% No Advocates say formal learning matters but “rarely, if ever, provides the complete answer”; some executives named coursework as pivotal Keep courses, and add follow-up that applies them
Fund manager coaching As a prompt only The “20” points to people and relationships, but no number says how much Tie coaching to a behavior you can observe
Add job aids and stretch assignments As a prompt only The “70” points to work, not to a specific support Pilot on one skill and compare results
Explain why a course alone will not change performance Yes This is the idea the model was built to carry Pair it with a transfer plan
Report the mix as a KPI No Share of hours says nothing about behavior or results Behavior and business measures

Swap The Percentage For A Question

When a slide says “70-20-10,” replace it with three questions for one named skill: where will people first use it, who will see them use it, and what feedback will they get? If you can answer all three, you have used the idea without needing the numbers.

How do you audit your own learning mix?

Audit by evidence for each channel, one role or skill at a time, rather than by percentage. Ask whether the work gives people practice, whether someone observes and coaches them, and whether formal learning ties to real tasks and gets followed up. A rating of strong, patchy or absent for each row is enough.

This audit is original to this guide. It is built from the model’s three channels and is not a validated instrument, so treat the ratings as a conversation starter for your team.

Channel Audit question Evidence to pull Warning sign
Experience Does the job give people regular chances to use this skill within weeks of learning it? Task assignments, rotation records, manager notes The skill is needed rarely or only under pressure
Experience Is there a job aid or reference at the point of need? Job aids, checklists, knowledge base searches People ask colleagues because nothing is written down
Experience Do people get feedback on how they did? Observation checklists, review comments Experience happens, but nobody sees the result
Social Does the manager discuss the skill before and after training? Pre-course and follow-up conversation records Managers learn about training from the completion report
Social Is there a coach, mentor or peer group for this skill? Coaching logs, community activity Mentoring exists on paper but has no agenda
Formal Does the course practice real job tasks? Course scenarios compared with job tasks Content is generic knowledge with a quiz
Formal Is there follow-up after the course? Scheduled check-ins, reinforcement activity Learning ends at the completion certificate
All Can you observe the target behavior on the job? Manager observation, quality or output data Success is measured only by completions

Read the pattern, not the score. Strong formal learning with absent experience and social channels is a course-heavy mix, and transfer is likely to be your problem. Strong experience with no feedback may repeat poor habits. Strong social support with no chance to practice leaves people with advice they cannot use.

How do you support the 20 and the 70 without a formula?

Give managers a defined role in learning and give people places to practice. A manager conversation before and after a course, a practice assignment and a follow-up check cover most of what the “20” and “70” ask for. None of it needs a percentage.

LMSpedia’s transfer of training guide says the direct manager often predicts transfer better than course design does, and recommends follow-up checks at 30, 60 and 90 days. For the social channel, Bandura’s social learning theory explains why observing capable colleagues matters. For coaching, the comparison of ILT and coaching formats suggests running instructor-led training first for a shared baseline and then adding coaching to help people apply it.

The on-the-job support is usually smaller than people expect:

  • Job aids: a one-page checklist or short video at the point of need.
  • Stretch tasks: a real assignment that uses the new skill, with a named reviewer.
  • Peer review: a colleague who has done the task reviews the first attempt.

Start With One Manager Conversation

Before building anything new, give managers a three-question script for the day a direct report finishes a course: what will you try first, when will you try it, and when should we talk about how it went? It takes minutes and puts the manager into the learning loop, which completion data never shows.

How do you measure whether your learning mix is working?

Measure behavior change and business results for the target skill, not the share of hours spent in each channel. If people do the task differently and the outcome improves, the mix is working whatever the ratio. If they finish courses but the work is unchanged, no percentage will fix it.

The Kirkpatrick model gives you the vocabulary: levels 3 and 4 ask about behavior and results. LMSpedia’s guide to learning and development KPIs argues for a small set of three to five headline measures that show whether training changes behavior, rather than activity counts such as completion rates. Pick the two or three that match the skill you audited above.

When should you retire the 70-20-10 percentages and keep the idea?

Retire the percentages whenever a number would become a target, a budget line or a KPI. Keep the idea whenever it widens the conversation beyond courses. As of November 2026, the evidence supports the second use and not the first.

Here is the practical split:

  • Keep: the reminder that much development happens through work and other people, as executives reported in the CCL study.
  • Keep: the argument that a course alone rarely changes performance.
  • Retire: any percentage in a budget, time target or scorecard.
  • Retire: any claim that research proves the 70, 20 and 10 shares.
  • Retire: applying the split to roles, such as frontline or regulated work, that the original sample did not cover, until you have your own evidence.

My verdict: treat 70-20-10 as a question to ask, not an answer to implement. The case for the percentages rests on a retrospective survey of successful executives. Critics found the evidence thin, and its advocates say the ratio varies. If your leadership likes the label, use it as a shorthand for “where will people practice, and who will help them,” and attach your own measures.

What should you do next?

Pick one role or skill that matters to the business and run the audit table with its manager. Rate each row as strong, patchy or absent, then choose the one weakest channel to fix first.

Next, draft a one-paragraph answer for the person who asked you to apply 70-20-10. Say that the numbers describe where executives said they learned. Say that the model’s own advocates call it a reference model and not a recipe. Then propose to improve practice and follow-up for one skill and measure behavior change.

Finally, check your existing plan for percentages. Wherever one appears as a target, replace it with a behavior you can observe.

FAQ

Q1. What is the 70-20-10 model for learning?

It is a reference model saying people develop mostly through work experience (70), through working with and learning from others (20), and through formal courses (10). Its advocates describe the numbers as a reminder of where learning happens, not a fixed split, and say the ratio varies with the work and the results needed.

Q2. Who created the 70-20-10 model?

It is usually credited to Morgan McCall, Michael Lombardo and Robert Eichinger of the Center for Creative Leadership. Their 1988 book Lessons of Experience studied 191 executives. According to ATD, the 70:20:10 phrase does not appear in that book, and Lombardo and Eichinger summarized the findings later in a 1996 planning tool.

Q3. Is the 70-20-10 model backed by research?

Not for the exact percentages, based on the sources reviewed here. The original work was a study of executives’ recalled career events. A 2012 paper, as quoted by ATD, found a lack of empirical data, and a 2018 review by Alan Clardy concluded the 70% rule of informal learning needs to be set aside.

Q4. Can you use 70-20-10 to split a training budget?

It is not designed for that. The 70:20:10 Institute presents the numbers as sources of learning, and its model page does not mention budgets. Treat any budget split as your own assumption. Fund by the performance gap you need to close, then check whether behavior and results change.

Q5. Is the 70-20-10 model still relevant in 2026?

The idea is still useful: development happens through work and other people, and a course alone rarely changes performance. The fixed percentages are not. As of November 2026, use the model as a prompt for questions about practice, feedback and follow-up, and keep the numbers out of targets and scorecards.

Q6. Does formal training still matter under the 70-20-10 model?

Yes. A post on Charles Jennings’ blog says formal learning “rarely, if ever, provides the complete answer,” which means it is part of the answer. In the executive study, some leaders also named coursework as a difference-making event, according to ATD. Design courses around real tasks and add manager follow-up so the learning transfers to the job.

Devendra Devendra

Written by Devendra Devendra

Bio Here

Table of contents