The 70-20-10 model is a design prompt, not a budget rule

The claim everyone quotes

Ask a room of learning and development people how adults learn at work, and someone will mention 70-20-10. The usual version: about 70% of what people learn comes from doing the job, 20% from other people such as a manager or a mentor, and 10% from formal courses and reading.

It is easy to see why it spread. It matches what many of us feel about our own careers: the hard project taught us more than the workshop did. It is short, it sounds precise, and it gives a learning team a simple picture to put on a slide.

And it gets used for more than a picture. Writing for ATD in 2014, Andrew Jefferson and Roy Pollock pointed to a blog that suggested training departments should spend only 10% of their time and resources on formal training. Once a ratio turns into a budget, it is worth checking where it came from.

Where the numbers came from

The model is most often traced to research at the Center for Creative Leadership (CCL), written up by Morgan McCall, Michael Lombardo and Ann Morrison in The Lessons of Experience (1988). They wanted to understand what made executives successful. According to Jefferson and Pollock, they asked 191 successful executives to name at least three key events in their careers that changed the way they managed, and what they learned from each.

The 70-20-10 phrase doesn’t appear in that book. The split shows up later, in Lombardo and Robert Eichinger’s Career Architect Development Planner (1996), which summarized the lessons learned by successful managers as roughly 70% from tough jobs, 20% from people (mostly the boss) and 10% from courses and reading.

Look at what that measures. These were accounts of memorable events, given by successful executives looking back. Memorable is not the same as where the learning happened, and senior executives are not the same as a new nurse, a sales hire or a support team. Their answers tell us what stood out to the people who made it to the top. They can’t tell us how learning divides up, and the round numbers came from a summary of them, not from a measurement.

What the reviews found

Jefferson and Pollock called 70-20-10 “a conceptual or theoretical model based on retrospective musings by executives” and wrote that it is “neither a scientific fact nor a recipe for how best to develop people.” They also quoted a 2012 review by DeakinPrime, which found “a lack of empirical data supporting 70:20:10”.

In 2018, Alan Clardy of Towson University went further in Human Resource Development Review. He examined five bodies of research that seem to agree that 70% or more of work-based learning happens informally, and found the evidence for this “70% rule” weak, citing sloppy scholarship, inconsistent concepts and basic problems with how the research was done. He concluded that HR development policies relying on it are likely to be misleading and that the rule “needs to be set aside.” His title sums it up: a fact in search of evidence.

None of this says informal learning is small or unimportant. It says nobody has measured the split, and, as Jefferson and Pollock put it, the question of how much learning happened where “is impossible to test.” If that sounds familiar, it should: it is the same pattern as learning styles, which we looked at in “The future of learning is personal”. A tidy idea that feels true travels faster than the research behind it.

What holds up underneath

Take the numbers away and the intuition underneath is sound: people learn a great deal from doing real work, getting feedback and watching others. For a training team, the useful research question is narrower: what makes learning carry over into the job?

A meta-analysis by Brian Blume and colleagues pooled 89 empirical studies of training transfer, the use of what was learned back at work. It found positive links between transfer and factors such as motivation, cognitive ability, conscientiousness and a supportive work environment. Most of those links were stronger for open skills, such as leadership development, than for closed skills, such as learning a piece of software.

In 2012, Eduardo Salas and colleagues reviewed the science of training for Psychological Science in the Public Interest. Their summary of the research is that “training works” and that the way it is designed, delivered and implemented matters. They describe training as a systematic process, with steps before, during and after it, and the Association for Psychological Science summed up their point: what happens before and after training is just as important for a program’s success as what happens during it.

Our view: a design prompt, not a budget rule

This part is our opinion, not a finding. We think 70-20-10 is worth keeping as a question and worth dropping as a number. As a question it is useful: for this program, where will people practice, who will give them feedback, and what should the formal part do? As a number, it invites mistakes.

We read the research the other way from the budget version. Good formal learning is what makes on-the-job learning work: it gives people a model to practice, words for the feedback and a reason for the manager to get involved. Even The Lessons of Experience, as Jefferson and Pollock point out, found that the coursework that mattered to executives “dealt with a relevant issue, and it occurred at a good time for the manager.” So design every program with practice and coaching built in, and stop using the ratio to cut the course. The ratio tends to invite four mistakes:

• Cutting courses to hit 10%, as if the course were the easy part to drop instead of the part that sets up everything else.
• Assuming learning “happens anyway” on the job, so nobody designs the practice, picks the stretch tasks or makes time for feedback.
• Treating the three parts as separate budgets owned by different teams, when the same skill should run through all three.
• Using the numbers to defend a plan, instead of asking what this group of learners needs.

A blended program, step by step

Here is a simple pattern we like for a skill such as running a first one-to-one meeting as a new team lead. It isn’t the only way, but each step has a job, and none of them is left to chance.

The course sets up the practice, the practice gives the coaching something real to talk about, and the quiz shows whether it lasted. That is closer to what the research describes, before, during and after, than any split by percentage.

How Classbase is designed to help

We are building Classbase for programs like this. A program in Classbase is a path of several courses, in sections that open on time: after the last one, on a date, or days after enrolling. Cohorts bring a group together for live sessions on Zoom, Teams or Meet, with attendance that counts toward completion. A follow-up test weeks later shows what stayed, and a short daily review brings questions back over the weeks that follow.

Between sessions, the AI tutor in each course answers from your own material and coaches practice conversations with hints instead of answers, graded on your criteria. It is a place to rehearse, not a replacement for the manager. The real assignment and the coaching talk still happen at work, between people, and we are working on bringing more of that part into the same place.

If you are rethinking how your academy splits its effort, book a demo and we’ll show you what a blended program looks like in Classbase.