70-20-10 Done Right: Building a Learning System That Actually Changes Behavior

The 70-20-10 model came out of decades of research at the Center for Creative Leadership (Morgan McCall, Robert Eichinger, Michael Lombardo) studying how successful executives actually grew. Their finding: roughly 70% of meaningful development came from challenging on-the-job experiences, 20% from people (bosses, mentors, peers) and only 10% from formal courses. The numbers were never meant to be a budget formula — they were a reminder that classroom training is the weakest of three legs.

Where 70-20-10 actually comes from

The model is grounded in The Lessons of Experience (McCall, Lombardo, Morrison, 1988), which interviewed nearly 200 executives at six Fortune 500 firms. The Center for Creative Leadership later codified it. Subsequent reviews (CIPD, ATD, Bersin) keep finding the same pattern: behavior change is dominated by what you do at work, with whom, and under what feedback — not by what you sat through.

“Development is about doing, not about being told.”
— Morgan McCall, Center for Creative Leadership

Why most companies misread it

They treat it as a budget split, then quietly spend 90% on courses because courses are easier to procure.
They confuse 'experience' with 'business as usual' — repeating the same project ten times is repetition, not stretch.
They have no system for the 20: mentoring is informal, feedback is annual, and managers are untrained as coaches.
They measure completion of courses, not change in behavior or business outcomes.

An operating model for 70-20-10

The three loops of a real learning system

1
Loop 1 — Capability map
Define the 8–15 capabilities the business actually needs by role family. Avoid vague 'leadership skills' — be specific (e.g. 'runs a structured hiring loop', 'designs an A/B test with statistical power').
2
Loop 2 — Development plan per person
Each person has 2–3 capabilities they are deliberately growing this cycle, mapped to a stretch project (70), a coaching/mentor partner (20) and a course/book/cohort (10).
3
Loop 3 — Evidence
Quarterly check-ins on visible behavior change, not effort. Manager + the person + one observer. Tied back to the capability map.

The 70: designed stretch experiences

The most powerful developmental events in McCall's research were: leading a turnaround, starting something from scratch, scope/scale jumps, high-stakes presentations, working across boundaries, and personal hardships handled at work. None of these happen by accident at scale — you have to design them.

Engineered stretch — examples by level

Career stage	Stretch experience to engineer	What it builds
IC → Senior	Own a cross-team initiative with two dependencies	Influence without authority, written comms
Senior → Staff/Manager	Lead a hiring loop for your replacement, or run an incident review	Hiring judgement, blameless leadership
Manager → Director	Run a quarterly planning process end-to-end	Strategy translation, prioritization, narrative
Director → VP	Open a new market / function / region from scratch	Ambiguity, capital allocation, executive comms

Stretch is not abandonment

A stretch assignment without a safety net is a setup. Pair every stretch with an explicit sponsor, a 'help is welcome' contract, and a pre-mortem on what could go wrong.

The 20: relationships and feedback

The 20 is where most companies leave the most value on the table. Three lightweight systems compound dramatically.

Mentor matching — quarterly cohorts, two-page mentor charters, 6-session minimum. Don't 'assign' organically; structure it.
Coaching for managers — 6–10 hours per manager per year of professional coaching, even shared coach pools work.
Peer learning circles — small groups (4–6) of same-level peers, monthly, problem-of-the-month format.

“The single biggest predictor of professional growth was the quality of the boss you happened to have.”
— Center for Creative Leadership, longitudinal studies

The 10: formal learning that earns its slot

Formal learning has a job: it transfers concepts faster than experience and shortens trial-and-error. Use it where the cost of learning by doing is high (legal, safety, ethics, security) or where shared vocabulary matters (a single Drucker chapter for the leadership team beats 12 different blog posts).

Choosing formal learning that actually transfers

1
Pre-work
Reflection prompts the week before — 'what is the hardest version of this problem you face right now?'
2
Practice, not lecture
Case-based, role-played, written reps. Bloom's mastery research is unambiguous: practice with feedback beats lecture by a wide margin.
3
Application contract
Each learner leaves with one specific commitment to apply within 14 days, shared with their manager.
4
Manager debrief
Two weeks later — 'show me what you tried.' Without this loop, transfer drops below 15%.

Kirkpatrick + Phillips: proving it worked

Donald Kirkpatrick's four-level model (1959) and Jack Phillips's later ROI extension are still the working standard. The discipline is to insist on at least Level 3 (behavior) for any program above $25k spend.

Evaluation levels — what to measure

Level	What it measures	Method	When to use
1 — Reaction	Did they like it?	End-of-session survey, NPS	Always; cheap signal
2 — Learning	Did they learn it?	Pre/post assessment, demonstration	Skill-based programs
3 — Behavior	Are they doing it?	Manager observation, 360, work sample	Above-the-line investment
4 — Results	Did the business move?	Linked KPI vs. control / baseline	Strategic programs only
5 — ROI (Phillips)	Net financial return	Benefit minus cost in same units	Programs >1% of payroll

Budget, ratios and what good looks like

1.3–1.6%

of payroll spent on L&D

Median, ATD State of the Industry

34 hrs

of learning per employee / year

Top-quartile firms

$1,280

median L&D spend / employee

ATD 2023 benchmark

94%

of employees

would stay longer if invested in their learning (LinkedIn Learning Report)

A defensible budget split

If you're starting from zero, allocate roughly: 40% manager enablement and coaching, 30% cohort programs at key transitions (IC→manager, manager→director), 20% on-demand library + tuition reimbursement, 10% compliance/required. Revisit after one year with Level-3 evidence.

The 90-day L&D rebuild

Days 1–15: Capability mapping with the top 3 functions. Output: 8–15 named capabilities per family.
Days 16–30: Manager-as-coach training for all people managers. One half-day workshop + practice partners.
Days 31–45: Define 3 strategic stretch experiences per function. Identify candidates and sponsors.
Days 46–60: Launch mentor matching cohort + peer circles. Cancel any course without a Level-3 plan.
Days 61–75: Build 1-page learning plans with every employee, signed by their manager.
Days 76–90: First quarterly evidence review. Publish one anonymized case study of behavior change.

Anti-patterns

Catalog thinking — 'we bought 14,000 courses' is not a strategy.
Heroes-only mentoring — only the visible top performers get sponsors; pipelines stay narrow.
Annual training week — learning treated as an event, not a system.
No manager loop — learners apply nothing because nobody asks.
Vanity metrics — hours consumed, NPS of training, 'engagement' with no behavior tie.

Sources to go deeper

References

The Lessons of Experience — McCall, Lombardo, Morrison — Center for Creative Leadership
Kirkpatrick's Four Levels of Training Evaluation — Kirkpatrick Partners
ATD State of the Industry — Association for Talent Development
LinkedIn Workplace Learning Report — LinkedIn
CIPD on 70:20:10 — CIPD
Peak — Anders Ericsson — On deliberate practice
Bersin: High-Impact Learning Organization — Josh Bersin Company

Written by Pawan Joshi. Sources cited inline. Last updated 2026-05-17.