Decisions are made at the margin. The question is never "should we do this at all?" in the abstract — it is "should we do one more unit?" Marginal benefit is the extra benefit from one more unit of an activity; marginal cost is the extra cost. Expand the activity until marginal benefit equals marginal cost. Beyond that point, the next unit costs more than it is worth. That equality is the optimal stopping rule. It applies to production (how much to make), consumption (how much to buy), hiring (how many to add), and any choice where quantity is variable.
The framework assumes you can measure or estimate the incremental effect of one more unit. In practice, marginal cost often rises (diminishing returns to the variable input) and marginal benefit often falls (the next unit of consumption or output is worth less). Where the two curves cross, you stop. Firms produce until marginal cost equals marginal revenue (the marginal benefit of selling one more unit). Consumers buy until marginal utility per dollar is equal across goods. The same logic applies to time allocation, feature scope, and capital deployment.
The strategic implication is that past costs and average costs are irrelevant to the next decision. What matters is the cost and benefit of the next unit. A sunk cost is gone; it should not affect whether you do one more. Average cost can be high while marginal cost is low — or the reverse. The discipline is to ask: what do we gain from one more, and what do we give up? When marginal benefit falls below marginal cost, stop. When it is still above, continue. The margin is where the decision lives.
The rule applies across domains. In pricing, the firm sets output where marginal revenue equals marginal cost; in consumption, the individual equates marginal utility per dollar across goods. In time allocation, the next hour goes to the activity with the highest marginal value. In capital allocation, the next dollar goes where expected marginal return is highest. The same logic underlies supply curves (marginal cost) and demand curves (marginal benefit). Equilibrium is the point where the two meet. Master the margin and you have a single framework for quantity decisions.
A common error is to average instead of marginalise. "Our cost per unit is $50" does not tell you whether the next unit is worth producing — that depends on the marginal cost of the next unit and the marginal revenue it brings. "We've already invested $2M" does not tell you whether to invest $100K more — that depends only on what the next $100K yields. Retraining the question onto the next unit is the core of the discipline.
Section 2
How to See It
Marginal analysis reveals itself whenever the right question is "one more or not?" Look for situations where quantity is variable and the next unit has a measurable or estimable cost and benefit. The pattern: we are at some level of activity; should we add one unit? Compare the incremental gain to the incremental cost.
Business
You're seeing Marginal Cost/Benefit when a founder asks whether to hire one more engineer. The marginal benefit is the extra output (features, speed, quality) that engineer would add. The marginal cost is salary, equity, and the coordination cost of a larger team. The decision is at the margin: if marginal benefit exceeds marginal cost, hire; if not, stop. Past hires and average cost per engineer are irrelevant to this decision.
Technology
You're seeing Marginal Cost/Benefit when a product team debates adding another option to a pricing page. The marginal benefit is the revenue from customers who would convert at that tier. The marginal cost is complexity, support burden, and cannibalisation of higher tiers. The optimal number of tiers is where the marginal benefit of one more tier equals the marginal cost. Not "how many tiers do we have?" but "does the next tier pay for itself?"
Investing
You're seeing Marginal Cost/Benefit when an investor decides whether to add to a position. The marginal benefit is the expected return on the next dollar deployed in that idea. The marginal cost is the opportunity cost of not deploying that dollar elsewhere (or holding cash). Position sizing is marginal analysis: add until the marginal expected return of the next dollar in this position equals the marginal return available elsewhere.
Markets
You're seeing Marginal Cost/Benefit when a consumer decides whether to buy one more unit of a good. The marginal benefit is the utility of that unit; the marginal cost is the price (and the forgone spending on other goods). Consumption adjusts until marginal utility per dollar is equal across goods. The same logic applies at the market level: supply reflects marginal cost; demand reflects marginal benefit; equilibrium is where they meet.
Section 3
How to Use It
Decision filter
"Before expanding or contracting any activity, ask: what is the benefit of one more unit, and what is the cost of one more unit? If marginal benefit exceeds marginal cost, do one more. If marginal cost exceeds marginal benefit, do one less. Ignore sunk costs and average costs. Optimise at the margin."
As a founder
Use marginal analysis for hiring, feature scope, and spend. The next hire should be justified by marginal product exceeding marginal cost (salary, dilution, coordination). The next feature should be justified by marginal user value or revenue exceeding marginal development and maintenance cost. The mistake is deciding by totals or averages — "we need a full team" or "our average cost per customer is X." The right question is whether the next unit pays for itself. When marginal benefit is hard to measure, approximate. When marginal cost is rising (diminishing returns), expect that the optimal scale is finite.
As an investor
Size positions at the margin. The next dollar in a position should have an expected return that matches the opportunity cost of capital (or the next best idea). When conviction is high, marginal benefit of the next dollar is high; you add. When the position is large or the idea is fully priced, marginal benefit falls; you stop. Ignore average cost basis when deciding to add or trim — that is sunk. The decision is always: from here, does one more unit of exposure improve expected outcomes?
As a decision-maker
Apply marginal thinking to time, budget, and scope. The next hour on a task has a marginal benefit (progress) and a marginal cost (forgone alternative use of that hour). Allocate time until marginal value is equal across activities. The next dollar in a budget line has a marginal benefit; compare it to the marginal benefit of the next dollar elsewhere. Scope creep is often a failure of marginal analysis — the next feature was not worth its marginal cost, but no one asked the question.
Common misapplication: Using average cost or total cost instead of marginal cost. "We've already spent $2M on this project" is irrelevant to whether the next $100K is worth it. The marginal cost of the next step is $100K; the marginal benefit is what that step delivers. Sunk costs should not drag you forward; marginal analysis can tell you to stop even when average cost is high.
Second misapplication: Ignoring that marginal cost and benefit change with scale. The first unit might have high marginal benefit and low marginal cost; the tenth might have low marginal benefit and high marginal cost (e.g. diminishing returns). The optimal quantity is where they cross. Assuming constant marginal cost or benefit leads to over-expansion or under-expansion. Re-estimate at the margin as you scale.
Bezos has repeatedly framed decisions at the margin: "What's the marginal cost of one more customer? One more SKU? One more geography?" Amazon's expansion into new categories and countries was evaluated incrementally — does the next unit of expansion generate marginal return above marginal cost? The same logic applied to Prime: the marginal cost of adding one more benefit (e.g. video) to Prime was weighed against the marginal benefit in retention and sign-ups. Decisions were made one unit at a time, not in bulk.
Buffett sizes positions and acquisitions at the margin. The next dollar of capital goes where expected marginal return is highest. He has said that the most important word in investing is "margin" — of safety, and of incremental return. When evaluating an acquisition, the question is not "is this company good?" but "what do we give up for the next dollar we put here?" Opportunity cost is the marginal cost of capital; the marginal benefit is the return on the next dollar deployed in that opportunity.
Section 6
Visual Explanation
Marginal Cost/Benefit — Optimal quantity is where marginal benefit equals marginal cost. To the left, expand; to the right, contract. Sunk and average costs do not determine this point.
Section 7
Connected Models
Marginal cost/benefit is the decision rule that ties production (supply), consumption (demand), and allocation together. It is the micro-foundation for equilibrium and the practical rule for "how much?" The models below either reinforce it (thinking at the margin, opportunity cost), extend it to markets (supply and demand, elasticity), or highlight the errors that undermine it (sunk costs, vague trade-offs). The models below either feed into it (opportunity cost, thinking at the margin), extend it (supply and demand, elasticity), or warn against misapplying it (sunk costs).
Reinforces
Thinking at the Margin
Thinking at the margin is the habit of asking "what happens if we do one more?" Marginal cost/benefit is the formalisation: the next unit has a cost and a benefit; compare them. The two are the same discipline. Thinking at the margin avoids all-or-nothing framing; marginal cost/benefit gives the stopping rule (expand until MC = MB).
Reinforces
Opportunity Cost
The marginal cost of doing one more unit of X is often the forgone benefit of the next best use of that resource — the opportunity cost. Marginal cost/benefit and opportunity cost are twin concepts: opportunity cost is what you give up; marginal cost is the incremental cost, which often includes that forgone benefit. Both focus the decision on the next unit, not the past.
Leads-to
Supply and Demand
Supply curves are marginal cost curves (at least in competitive markets). Demand curves reflect marginal benefit (willingness to pay). Equilibrium is where supply meets demand — where marginal cost equals marginal benefit at the market level. Marginal analysis is the micro-foundation for supply and demand.
Leads-to
[Elasticity](/mental-models/elasticity)
Elasticity measures how quantity responds to price. The responsiveness of quantity to price depends on how steep the marginal benefit and marginal cost curves are. Inelastic demand means quantity does not change much when price changes — marginal benefit is steep. The link between elasticity and marginal analysis is direct: both describe the shape of the curves that determine optimal quantity and price.
Section 8
One Key Quote
"The marginal utility of a thing to anyone diminishes with every increase in the amount of it he already has."
— Alfred Marshall, Principles of Economics (1890)
Marshall was stating the law of diminishing marginal utility — the next unit is worth less. That declining marginal benefit, combined with rising marginal cost (from diminishing returns in production), ensures an interior optimum. You do not consume or produce infinitely; you stop where the next unit is no longer worth its cost. The quote captures one side of the margin; the full decision rule is to equate both sides.
Section 9
Analyst's Take
Faster Than Normal — Editorial View
Most decisions are marginal, not binary. The question is rarely "should we do this?" It is "should we do one more?" One more hire, one more feature, one more dollar of spend, one more hour of effort. Train the habit: what is the marginal benefit and marginal cost of the next unit? When they are equal, stop. When benefit exceeds cost, expand.
Sunk costs are the enemy of marginal thinking. The money or time already spent does not affect the marginal cost of the next step. The marginal cost of continuing is only what the next step costs. Founders who say "we've come this far" are applying sunk cost logic. The correct question is: from here, does one more unit pay for itself? If not, stop regardless of how much is sunk.
Average cost and marginal cost diverge. A project can have high average cost (total cost / quantity) but low marginal cost (cost of one more unit). In that case, continuing can be rational even when the project is "losing money" on average — you are recovering some fixed cost and the next unit has positive margin. The reverse also holds: low average cost does not imply that the next unit is worth doing. Always reason from the margin.
Use it for scope and prioritisation. The next feature has a marginal benefit (user value, retention, revenue) and a marginal cost (build time, complexity, support). Rank by marginal benefit per unit of marginal cost — or add until marginal benefit equals marginal cost. The same logic applies to backlog: the next item to do is the one with highest marginal value per unit of marginal effort. Prioritisation is marginal analysis.
Estimate when you cannot measure. Perfect marginal cost and benefit are rarely observable. Approximate. What would one more hire likely add? What would one more tier or feature likely add? Order-of-magnitude estimates at the margin beat precise totals that ignore the margin. The goal is to get the direction right: are we past the optimum or not?
Revisit the margin as conditions change. Marginal cost and benefit shift with technology, competition, and scale. What was optimal last quarter may not be optimal now. The decision to add or cut should be re-evaluated when the environment changes. Fixed rules ("we always add one more sales rep per $X of pipeline") can drift away from the true optimum. Re-anchor on the margin periodically.
Section 10
Test Yourself
Is this mental model at work here?
Scenario 1
A company has spent $5M on a product that has not launched. The team says one more $500K will get it to market. Expected revenue is $800K in year one.
Scenario 2
A consumer has ten streaming subscriptions and is considering an eleventh. She values the new one at $6/month. The subscription costs $10/month.
Scenario 3
A factory is losing money on average but keeps running. Price is above variable cost per unit but below full cost per unit.
Scenario 4
An investor adds to a position after the stock has risen 50%. Her cost basis goes up.
Summary. Marginal cost/benefit is the rule that decisions are made at the margin: expand until the benefit of one more unit equals the cost of one more unit. Ignore sunk and average costs; only the next unit matters. Use it for hiring, scope, spend, and allocation. The optimum is where marginal cost equals marginal benefit. Master the margin and you have a single framework for quantity decisions.
Section 11
Top Resources
Marginal analysis is core to microeconomics and decision-making. Marshall is the foundational source; modern texts and strategy books apply it to firms and choices.
Marshall's synthesis of marginal utility, marginal cost, and the equilibrium where supply (marginal cost) meets demand (marginal benefit). The source for the formal structure of marginal analysis and its role in price and quantity determination.
Standard treatment of marginal cost, marginal revenue, and profit maximisation. Clear link between production (diminishing returns, cost curves) and the marginal decision rule. Good for tying the math to intuition.
Kahneman documents systematic failures to think at the margin — sunk cost fallacy, framing effects. Understanding these biases reinforces why marginal analysis is a discipline: the default is to reason from totals and past cost. The book is a complement to the economic rule.
Munger advocates "elementary, worldly wisdom" including marginal thinking. His treatment of opportunity cost and incremental reasoning in business and investing. Practical application of marginal logic to capital allocation and life decisions.
Coase asked why firms exist instead of only markets. Part of the answer is that marginal cost of transacting (contracting, monitoring) can exceed the marginal benefit at some scale — so the firm internalises up to that point. Marginal analysis applied to the boundary of the firm.
Buffett's letters repeatedly illustrate marginal and opportunity-cost reasoning: the next dollar of capital goes where expected return is highest; past cost basis is irrelevant to the add/trim decision. Practical application of marginal analysis to capital allocation and position sizing.
Tension
Sunk Costs (Economics)
Sunk costs are irrelevant to the marginal decision. But psychologically, people let sunk costs drag them forward — "we've already invested so much." Marginal cost/benefit says: ignore sunk cost; only the cost and benefit of the next unit matter. The tension is between the correct rule (marginal) and the common error (sunk cost fallacy). Discipline is required to apply marginal analysis when sunk costs loom large.
Tension
[Trade-offs](/mental-models/trade-offs)
Trade-offs say that more of one thing means less of another. Marginal cost/benefit says how much more: expand until the marginal benefit of one more unit equals the marginal cost. Trade-offs define the frontier; marginal analysis finds the best point on that frontier. The tension: trade-offs are often framed in broad terms ("quality vs speed"). Marginal analysis forces the question: what is the marginal gain from one more unit of quality, and what is the marginal cost in speed?
