Scalability of value

How much value can one person’s work create, from a cook to a founder coach, a VC, or the founder of Amazon?

This single table compares 85 roles, ordered by their modeled annual benefit. It includes value delivered directly and value that travels through a few clients to the people their work affects.

These are illustrative scenarios, not measured occupational averages. For each row:

Annual benefit = people or services supported per year × assumed benefit per person or service.

For a coach or adviser, the per-client estimate includes a hypothetical downstream effect: baseline annual benefit of the client’s work or system × assumed improvement caused by the support. The calculation appears in the same column as a cook’s benefit per meal. A few clients can therefore produce a large total in this model.

Reach counts meals, visits, or trips for recurring services, and people or clients otherwise. Dollar amounts represent assumed benefit over the next-best alternative, not pay, revenue, profit, net worth, or personal worth. Founder rows describe an organization at the stated scale. Depth is a separate subjective score from 1 (small convenience) to 5 (potentially life-changing); roles without a depth estimate are marked “Not scored.” Overlapping rows should not be added together.

On a phone, scroll the table sideways to see the value columns.

The ordering follows the assumptions, not a claim that one occupation is inherently worth more than another. For example, the cook scenario is 20,000 meals × $3 = $60,000 per year. The founder-coach scenario is 20 clients × $10 million of annual benefit per client’s system × 1% improvement = $2 million per year. For the coach, that entire estimate depends on the assumed improvement actually reaching useful outcomes.

A small change at a large organization is not automatically easy to achieve. With no causal effect, an adviser’s indirect benefit is zero; harmful advice can make it negative. Several helpers may contribute to the same improvement, and each cannot claim the entire result. The direct health or relationship benefit to a client may also matter even when no work-related effect is established.

Reading the numbers

The cook scenario is 20,000 meals × $3 of benefit per meal = $60,000 per year. Cooking another meal still requires ingredients, equipment, and someone’s work.

The hotel scenario is 20,000 guest-nights × $30 of benefit = $600,000 per year at the hotel level. The owner’s leverage comes from coordinating a building and a team. The model does not assign all of their contribution to the owner.

For the software engineer, 100,000 users × $20 of additional annual benefit = $2 million. That requires an improvement that actually helps those users. Merely touching code used by 100,000 people doesn’t establish that benefit.

The Amazon-founder scenario is 300 million customers × $100 of annual benefit = $30 billion at the organization level. Here, 300 million and $100 are model inputs, not verified customer counts or measured consumer savings. At $10 per customer, the same calculation gives $3 billion; at $200, it gives $60 billion. The estimate is sensitive to what the product changes relative to customers’ alternatives.

For a factual sense of company scale, Amazon reported $716.9 billion in 2025 net sales in its 2025 results. Sales measure money paid to the company. They do not establish the benefit above, or how much of it Jeff Bezos personally caused. Employees, sellers, suppliers, investors, and customers all contribute to the system.

Duration: how much survives after you stop?

A meal is consumed. The same recipe can be used again. A book, piece of software, institution, or discovery can keep producing benefits after its creator stops working on it, although distribution and maintenance may still take work.

My original shorthand was:

Reach, depth, and duration. In the dollar table, depth is represented by the assumed benefit per person or service, and the time window is one year. To estimate several years, I would add each year’s benefit, allowing for adoption, decay, ongoing costs, and harms. Multiplying by years only works if the annual benefit stays constant.

Following the effect through another person

There is another route to scale: help someone whose own work reaches many people.

Suppose a coach works with 20 founders, and each founder’s organization produces $10 million of annual benefit under the same kind of model. If coaching causes a 1% improvement in that benefit, the indirect effect is:

That is a hypothetical causal improvement in useful outcomes, not a 1% change in mood automatically passed through to everyone. With no effect on those outcomes, the modeled indirect benefit is zero. With overlapping organizations or customers, I would also need to remove double counting.

This is why a role with very few direct clients can still have a large indirect reach. For a VC, the question is what useful work changes because of their selection, funding, introductions, or support compared with the funding and help the founder could otherwise obtain. A check’s size or a fund’s return alone does not answer that question. The same reasoning applies to helping investors, scientists, policymakers, or other coaches. With each additional link, I would need evidence that the improvement actually reached the next person. A longer chain creates more opportunities for both compounding and loss.

What the table leaves open

The health and relationship numbers are especially subjective. A dollar equivalent cannot fully describe being cared for or having someone close to you. Some benefits cannot be bought at the stated price, and a high score here does not rank a person’s worth.

The table models gross benefit under favorable assumptions. Net value also needs the resources consumed, harms, and what would have happened anyway. A globally distributed product that makes people’s lives worse can scale negative value just as effectively.