marginal cost
Marginal cost answers a single, very practical question: what does it cost to make just one more? Not the average cost, not the total — the extra cost of the next unit. A bakery already baking 200 loaves wonders what the 201st adds to its bill. That little extra is the marginal cost, and it is the most decision-relevant cost there is.
Precisely, marginal cost is the change in total cost when output rises by one unit. If baking 200 loaves costs 600 and baking 201 costs 602.50, the marginal cost of that loaf is 2.50. Notice that fixed costs never enter: the rent is paid anyway, so it adds nothing to the cost of one more loaf — marginal cost comes purely from extra variable inputs. And because of diminishing returns, marginal cost is usually U-shaped: it falls at first (early workers specialise and run smoothly) then rises (later workers crowd the fixed equipment, so each extra loaf eats more inputs). There is a tidy mirror here — marginal cost is low exactly when marginal product is high, and vice versa.
Marginal cost is the single most important cost concept in economics because firms decide at the margin. The profit-maximising rule is to keep producing as long as the revenue from one more unit (marginal revenue) is at least its marginal cost, and to stop where marginal revenue equals marginal cost. The rising part of the marginal cost curve is, for a competitive firm, literally its supply curve — it shows how much it will offer at each price. Marginal cost is the link between the firm's recipe and the market's prices.
An airline's flight is going anyway with one empty seat. Filling it costs almost nothing extra — a little fuel, one more snack — so the marginal cost of that passenger is tiny, even though the average cost per seat (which shares the plane, crew and fuel) is far higher. That gap is why last-minute fares can be so cheap.
Marginal cost is the price tag on the very next unit — fixed costs don't count.
Marginal cost ignores fixed costs entirely — that is the point, not an oversight. Basing the next production decision on average cost instead is a classic mistake; what matters for one more unit is only the extra cost it actually causes.