How Founder Decision-Making Changes as a Company Grows

In a company of five people, the founder makes nearly every decision, and that's usually right. The founder has the context, the customers are a Slack message away, and a wrong call is cheap to reverse. At fifty people the same habit starts to cost real time. At five hundred it becomes the main constraint on how fast the company can move.

The shift isn't about making better decisions. It's about changing what the founder's job is: from deciding to designing how decisions get made. That means who decides what, with whose input, and with what review. This article lays out the decision load at different sizes, then walks through four named models that give the shift a vocabulary: Larry Greiner's growth phases, Jeff Bezos's one-way and two-way doors, Bain's RAPID framework, and decision rights matrices. It ends with the bottleneck effect and a practical sequence for handing decisions off.

If you want the broader pattern behind this, see founder dependence and founder's syndrome.

Why the decision load changes with scale

Two things grow as a company does, and they grow at different speeds.

The number of decisions grows fast. Every new hire, customer segment, product line, and market adds decisions: pricing exceptions, hiring offers, vendor choices, scope trade-offs, escalations. The count rises much faster than headcount because decisions multiply across the connections between people and teams.

The founder's capacity doesn't grow at all. There are still the same hours in the week. The founder's attention is the fixed resource, and every decision routed through them spends some of it.

At small scale, the founder is also the best-informed person for nearly every question. That's what makes centralization efficient early on. As the company grows, the people closest to a decision, a sales rep with a discount request or a support lead with a refund case, increasingly know more than the founder does about that specific situation. The founder's context advantage shrinks, but the habit of routing everything upward stays.

That's the root of the problem. The habit was correct at one size and is wrong at another, and nothing in the daily experience signals when the line is crossed.

Greiner: the style that works becomes the crisis

Larry Greiner's article "Evolution and Revolution as Organizations Grow" first appeared in Harvard Business Review in July-August 1972, and HBR republished it as a Classic in 1998. The HBR summary states the core idea in one line: management practices that work well in one phase may bring on a crisis in another.

Greiner described growth as alternating periods of calm and crisis. Each calm period runs on a dominant management style, and each ends in a predictable crisis that the old style can't solve. The early phases line up closely with how founder decision-making changes. This summary follows a secondary explainer of the model.

Greiner's later phases (coordination, then collaboration) cover large organizations and are less relevant to a founder's first few hundred hires. The first two crises are the ones that matter here, and they map onto two different founder mistakes.

The crisis of leadership is the moment the founder's improvised, personal decision-making stops scaling. The fix isn't more effort. It's installing a structure and often a management layer. Management team below the founder covers that build.

The crisis of autonomy is subtler. The company now has managers, but they don't have real decision authority, because the founder still approves anything that matters. Capable people feel they're running errands. Greiner's point is that the way out is delegation, and that delegation itself brings the next crisis, control. So the real task isn't to pick delegation or control. It's to design delegation with enough guardrails that control survives.

Bezos: one-way doors and two-way doors

The most useful single idea for sorting which decisions a founder should keep is from Jeff Bezos's 2015 letter to Amazon shareholders, filed with the SEC.

Bezos separates decisions by reversibility. Some decisions are consequential and irreversible or nearly irreversible, which he calls one-way doors. These, the letter says, must be made methodically, carefully, slowly, with great deliberation and consultation. He calls them Type 1 decisions. Most decisions, he says, aren't like that. They're changeable and reversible, which makes them two-way doors, or Type 2 decisions. According to the letter, Type 2 decisions can and should be made quickly by high-judgment individuals or small groups.

The letter then names the failure mode that matters for growing companies. As organizations get larger, it says, there seems to be a tendency to use the heavyweight Type 1 process on most decisions, including many Type 2 decisions. The result, in the letter's words, is slowness, unthoughtful risk aversion, failure to experiment sufficiently, and diminished invention.

For a founder, this is a practical sorting test:

The examples in the table are our illustration, not a list from the letter. What belongs in each column depends on the company. A $2,000 software purchase is a two-way door for a company with $50 million in revenue and a one-way door for one with $50,000.

The takeaway is that founders don't need to give up decisions in general. They need to separate the small number of Type 1 decisions they should stay in, from the large number of Type 2 decisions that should leave their desk. For the risk side of that sorting, risk-based decision making goes deeper.

Bain's RAPID: naming who does what in a decision

Sorting decisions by weight solves half the problem. The other half is clarity about who plays which role in each one. Bain & Company's RAPID framework names five roles: Recommend, Agree, Perform, Input, and Decide.

Two details in Bain's description matter a great deal for founders. First, the decider should sit as close to the implementation as possible, which is almost the opposite of the default in a founder-led company, where the decider is whoever is most senior. Second, Bain notes that not every decision needs explicit RAPID roles. It suggests starting with high-value or high-frequency decisions.

In founder-led companies, RAPID tends to expose a common pattern: the founder is listed as Decide on dozens of decisions that need only an Input from them. Moving the founder from Decide to Input (or to Agree where a real constraint exists) is often the single most useful change, and it costs nothing but discipline.

Decision rights and delegation matrices

RAPID works decision by decision. A decision rights matrix applies the same logic across a whole category of decisions, written down once so people don't need to ask.

The usual form is a table with decision types in rows and roles or levels in columns, plus limits: spending thresholds, discount floors, hiring bands, contract terms that need sign-off. A well-built matrix answers three questions without a meeting:

1. Who can decide this on their own?
2. Up to what limit?
3. Who must be told, or must approve, above that limit?

The practical value is that it converts "ask the founder" from a default into an exception. The full build, with examples and common mistakes, is in delegation of authority matrix.

A few rules of thumb hold up across companies:

• Set limits in numbers, not adjectives. "Reasonable discount" invites escalation. "Up to 15% without approval" doesn't.
• Write down what stays with the founder. A short list of retained decisions is easier to defend than silence, and it tells the team where the real lines are.
• Review it on a schedule. Limits that fit a $5 million company are wrong at $25 million.
• Expect some wrong calls. If nothing delegated ever goes wrong, the limits are too tight.

The bottleneck effect

When every decision routes through one person, the organization behaves like a queue. Decisions wait, and the cost of waiting rarely shows up on any dashboard.

Three effects follow, and they compound.

Latency. Each decision waits for the founder's calendar. Even a short wait, repeated across hundreds of decisions, adds up to weeks of lost motion. And the decisions that wait longest are often the small ones, because the founder rightly prioritizes the big ones.

Learning stops. People who never decide don't build judgment. They learn to prepare a question for the founder, not to answer it. That's how a company ends up with managers who can't act alone, which then justifies the founder staying involved. The loop feeds itself. The management team below the founder article covers how this pattern blocks a leadership layer from forming.

Risk concentrates. If decisions run through one person, so does the knowledge behind them. That ties directly to key person risk.

Bezos's letter describes the same pattern from the other direction: heavyweight process applied to reversible decisions produces slowness and risk aversion. A founder bottleneck is the same failure in a more personal form.

From making decisions to designing decision-making

Putting the models together gives a sequence that tracks how the founder's role shifts.

Stage 1: decide almost everything. In the early stage this is appropriate. The risk is that it becomes a habit that outlasts its usefulness. Greiner's first phase is exactly this style, and its crisis is the point where it stops working.

Stage 2: sort the decisions. Start separating Type 1 from Type 2. Make a short list of decisions that really are one-way doors for the company, and let everything else begin to move.

Stage 3: name the roles. For recurring, high-value decisions, use RAPID-style roles. Pick one recommender and one decider per decision, and make the decider someone close to the work.

Stage 4: write the rights down. Turn the repeated patterns into a delegation matrix with real limits. This is where Greiner's autonomy crisis gets resolved, because managers now have written authority and not just permission on a good day.

Stage 5: inspect, don't approve. The founder moves from approving individual decisions to reviewing patterns: a sample of decisions made, outcomes, and exceptions. The question changes from "should we do this?" to "is the system producing good decisions?"

This is the heart of the shift. Designing how decisions get made is itself a Type 1 task, because the structure sets the quality of thousands of later Type 2 calls. It's the highest-leverage decision work a founder does at scale.

The shift also changes what the founder should expect to feel. Once decisions move to others, some will be made differently than the founder would have made them. Many will be fine. A few will be wrong. Greiner's third-phase crisis of control is the cost of that discomfort: when authority is pushed down, leaders sense they've lost their grip. The answer in his model isn't to pull authority back, but to add coordination and review. Vroom-Yetton is a useful companion when the question is how much a leader should involve others in a given decision.

Key Facts: Founder decision-making at scale

• Larry Greiner's growth model first ran in HBR in July-August 1972; its core idea is that practices that work in one phase may bring on a crisis in another.
• In Greiner's model the early crises are leadership (phase 1) and autonomy (phase 2); delegation then leads to a crisis of control.
• Bezos's 2015 shareholder letter splits decisions into Type 1 (irreversible, one-way doors) and Type 2 (reversible, two-way doors) and says most are Type 2.
• The same letter says larger organizations tend to apply heavyweight Type 1 process to Type 2 decisions, producing slowness, risk aversion, and less experimentation.
• Bain's RAPID assigns five roles (Recommend, Agree, Perform, Input, Decide), with ideally one decider per decision, positioned close to implementation.
• Bain advises starting RAPID with high-value or high-frequency decisions, not every decision.

Related reading

• Delegation of Authority Matrix
• Founder Dependence
• Management Team Below the Founder
• Key Person Risk
• Vroom-Yetton Decision Model
• Risk-Based Decision Making
• Founder Mode vs Manager Mode Explained
• Common Founder Blind Spots