Picture the version of your team that exists eighteen months after your best person leaves. If the honest answer is "we'd struggle," you don't have a development gap, you have a single point of failure wearing a lanyard. Developing future leaders is the slow, deliberate work of making sure no role, including yours, depends on one irreplaceable person.

The quick version

  • People grow mostly through hard, real work, not courses. The common rule of thumb is 70-20-10: roughly 70% from stretching assignments, 20% from other people, 10% from formal training.
  • Each step up the ladder is a different job, not a bigger one. Promote someone without re-teaching them, and you keep your best individual contributor and lose them as a manager.
  • A successor is a named person being deliberately readied for a specific role, not a vague sense that "someone could probably step up."
  • The cheapest way to develop people is also the one most managers skip: give away the work that scares you to delegate.

The idea in depth

Three ideas do most of the heavy lifting here. The first is about how people actually learn to lead; the second is about what changes as they rise; the third is about how organisations track who is ready. Each is useful, and each has a failure mode worth naming out loud.

People learn to lead by doing leadership-shaped work

The best-known framework here is the 70-20-10 model, which grew out of research by Morgan McCall, Michael Lombardo and Robert Eichinger at the Center for Creative Leadership and was put into its familiar form by Lombardo and Eichinger in their 1996 Career Architect Development Planner. Drawing on surveys of executives reflecting on what had shaped them, the work found that the experiences that built leaders were overwhelmingly on-the-job: roughly 70% came from challenging assignments, 20% from other people (managers, mentors, peers), and only about 10% from formal courses.

The practical shift is to stop treating "development" as something that happens in a training room and start treating it as something you assign. If you want someone to grow, the most powerful thing you can hand them is a piece of real work that sits slightly beyond their current reach, a project to own, a difficult stakeholder to manage, a decision to make without you in the room. Delegation, used as a development tool rather than a way to offload tasks, is the single most repeatable lever you have.

The honest limitation: the 70-20-10 ratios are a memorable heuristic, not a measured law. They came largely from successful managers self-reporting what they thought had shaped them, a method that is prone to hindsight bias, and the model has been criticised for thin empirical validation of the exact split. Treat the numbers as a reminder of where the weight lies, not as a budget formula. The defensible claim is the direction (experience > relationships > courses), not the decimal places.

flowchart LR
    A(["Stretch
assignment ~70%"]) --> D(["Leadership
capability"])
    B(["Coaching &
mentoring ~20%"]) --> D
    C(["Formal
training ~10%"]) --> D
The 70-20-10 heuristic: most growth comes from real work, supported by people, topped up by training. Leaders Loop

Every promotion is a change of job, not a reward

The second idea comes from the Leadership Pipeline, set out by Ram Charan, Stephen Drotter and James Noel (first published in 2000, expanded in 2011). Their argument is that a career isn't a smooth ramp but a series of distinct passages, moving from managing yourself to managing others, then to managing managers, then to running a function, and so on. Each passage demands that the person learn new skills, spend their time differently, and value different things. The classic trap is the brilliant engineer promoted to manager who keeps doing the engineering, because that is what earned the promotion and what still feels like "real work."

The most expensive mistake in development is promoting your best individual contributor and assuming the new job is just the old one with more of it.

What helps is making the transition explicit. When you promote someone, name the work they need to stop doing as clearly as the work they need to start. Sit down in the first month and ask: what did you used to be measured on, and what are you measured on now? Where the old reflexes are strongest, solving instead of delegating, doing instead of coaching, that's where the development plan goes. This is also where coaching skills stop being a nice-to-have: a manager learning to ask rather than tell is a manager learning the new job.

The limitation: Charan's pipeline was built around large, layered corporations. In a flat startup or a small team, the passages blur and a single person may straddle three of them at once. Use it as a map of the kinds of change people go through, not a literal org chart you must replicate.

Knowing who is ready, and being honest about it

To plan succession you eventually need a shared, deliberate view of who could go where. The most common tool is the 9-box grid, which plots people on two axes, current performance against future potential, into nine cells. It was not born in HR at all: it began as the GE-McKinsey matrix in the 1970s for comparing business units, and was later repurposed to compare people. The appeal is that it forces a real conversation: a high performer in their current role is not automatically a high-potential for the next one, and the grid makes you say which is which.

The point is to run a calibration conversation, not to fill in a spreadsheet alone. Get the managers of a team in a room, place people on the grid together, and argue. The value is almost entirely in the disagreement, "you call her high-potential, I've seen her avoid every hard conversation", because that surfaces the development gaps no single manager sees.

The limitation is real and worth stating plainly: "potential" is slippery, easily confused with how much someone resembles the people already in charge, and a label written on a grid can quietly become a self-fulfilling prophecy. Used carelessly the 9-box bakes in bias and demotivates everyone parked in a low box. Treat it as a structured prompt for honest debate, reviewed regularly, never as a verdict, and never as something the person being boxed can't eventually see and discuss.

A worked example

Maya runs a 12-person customer-operations team. Her strongest performer, Daniel, resolves more cases than anyone and is the obvious answer when she imagines who'd run the team if she moved. The lazy version of "developing him" is to tell him he's next in line and send him on a two-day management course. (All figures here are illustrative.)

Instead, Maya works the three ideas. 9-box: in a calibration with her peer manager, she admits Daniel is a top performer but an unknown potential, nobody has seen him handle conflict or set direction. Pipeline: she names the passage out loud, "the job isn't closing more cases, it's getting the team to close them well without you." 70-20-10: she stops planning a course and starts assigning real work. Over the next quarter Daniel owns the weekly team huddle, leads the rollout of a new triage process (a project with genuine stakes and a real chance of going wrong), and mentors the two newest hires. Maya keeps a standing fortnightly 1:1 that is pure coaching, questions, not answers, plus a deliberate, escalating set of decisions she lets him make without her.

By the end of the quarter Maya has data she never had before: Daniel is superb at process and visibly uncomfortable with underperformance, avoiding the one hard conversation the rollout demanded. That's not a failure, it's the most useful thing the assignment produced. It tells her exactly where the next six months of development go, and it means that if she's hit by the proverbial bus, the team has a real, half-built successor rather than a name on a slide.

flowchart TD
    A(["Spot potential
(9-box calibration)"]) --> B(["Name the next
passage explicitly"])
    B --> C(["Assign stretching
real work"])
    C --> D(["Coach through it
in 1:1s"])
    D --> E(["Review what the
work revealed"])
    E --> A
Developing a successor is a loop, not a one-off nomination. Leaders Loop

Frequently asked questions

Isn't this just succession planning for big companies?

The board-level version, naming a CEO successor, is, but the discipline scales all the way down. Any time one person's departure would hurt, you have a succession question. For most managers the practical unit isn't "who runs the company" but "who could run my team," and that's answerable with the same three ideas at a fraction of the formality.

Won't developing my best people just make them leave faster?

The opposite is the more common outcome. People are far likelier to leave a role with no visible path than one where someone is deliberately stretching them. Yes, a developed person is more employable, but the alternative, hoarding growth to keep someone in place, tends to produce exactly the quiet, resentful exit you were trying to avoid.

What if I genuinely have no high-potential people to develop?

That's usually a sign you're measuring the wrong thing, confusing current performance with potential, or screening for people who already act like the leaders you have. Before concluding the bench is empty, give two or three people a real stretch assignment and watch. Potential is far easier to see in motion than in a review form.

How is this different from coaching or mentoring?

Coaching and mentoring are tools inside this work, not substitutes for it. Coaching helps a person think through a challenge they already own; mentoring lends them someone else's experience. Developing successors is the wider game of deciding who should grow toward what, then using coaching, mentoring and stretch work to get them there.

How long does it actually take to build a successor?

Longer than you'd like and shorter than never. Charan's pipeline framing is blunt about this: a real passage can't be crossed in a day or by taking a course. Plan in quarters and years, not weeks, which is exactly why starting before you need the successor, rather than during the crisis, is the whole point.

Related in the Toolkit

Where to go next