Approximating with Marginals — Using Marginals to Approximate Changes - Business Calculus

SECTION 1

Historical Context & Motivation

The concept of using derivatives to approximate small changes in economic quantities has roots stretching back to the birth of both calculus and modern economics. In the late nineteenth and early twentieth centuries, economists recognized that the tools Isaac Newton and Gottfried Leibniz had developed for physics could be powerfully repurposed for analyzing production costs, revenues, and profits. The idea was deceptively simple: rather than recomputing an entire cost or revenue function from scratch each time output changes, one could use the derivative — the instantaneous rate of change — to estimate the effect of producing one additional unit. This approximation technique became the backbone of marginal analysis, one of the most practical tools in managerial economics and business calculus.

1687 • Newton's Principia Published

Isaac Newton formalizes the calculus of fluxions, providing the mathematical foundation for rates of change that would later underpin marginal analysis in economics.

1871 • Marginalist Revolution Begins

William Stanley Jevons, Carl Menger, and Léon Walras independently develop marginal utility theory, applying differential calculus to economic decision-making for the first time.

1890 • Marshall's Principles of Economics

Alfred Marshall synthesizes marginal analysis into a comprehensive economic framework, introducing marginal cost and marginal revenue as central concepts in firm behavior.

1947 • Samuelson's Foundations

Paul Samuelson's Foundations of Economic Analysis rigorously establishes calculus-based optimization as the standard analytical method for economics, cementing the role of derivatives in business analysis.

The central question that marginal approximation addresses is practical and urgent: if a firm currently produces q units and is considering producing one more (or a few more), how much will its total cost, revenue, or profit change — without needing to evaluate the entire function from scratch? The derivative provides an elegant linear approximation that is both fast and surprisingly accurate for small changes in output, making it indispensable in real-time business decisions.

SECTION 2

Core Principles & Definitions

Before diving into calculations, it is essential to establish the foundational concepts that make marginal approximation work. The key insight is that a derivative evaluated at a particular production level gives us the slope of the tangent line to a cost, revenue, or profit curve at that point. For small changes in the independent variable (typically quantity), the tangent line closely tracks the actual curve, and so the derivative-based estimate closely approximates the true change. This principle extends naturally to the three primary business functions: marginal cost, marginal revenue, and marginal profit.

Marginal Cost — C′(q)

The derivative of the total cost function C(q). It approximates the additional cost of producing the (q + 1)th unit: ΔC ≈ C′(q) · Δq. When Δq = 1, marginal cost directly estimates the cost of one more unit.

Marginal Revenue — R′(q)

The derivative of the total revenue function R(q). It approximates the additional revenue earned from selling one more unit: ΔR ≈ R′(q) · Δq. This drives pricing and production decisions.

Marginal Profit — P′(q)

The derivative of the profit function P(q) = R(q) − C(q). Since differentiation is linear, P′(q) = R′(q) − C′(q). Marginal profit tells a firm whether producing one additional unit adds to or subtracts from total profit.

Linear Approximation Formula

The general tangent-line approximation: f(q + Δq) ≈ f(q) + f′(q) · Δq. This is the mathematical engine behind every marginal approximation. The smaller the Δq, the better the estimate.

✦

KEY TAKEAWAY

Think of marginal analysis like using a weather forecast to plan your afternoon. The forecast (the derivative) is based on current conditions (the function value at q) and tells you the trend (rate of change). For the next hour or two (small Δq), the forecast is very accurate. But if you try to predict weather a week out (large Δq), the forecast becomes unreliable. Similarly, marginal approximations are most accurate for small changes in quantity because the tangent line diverges from the actual curve as you move farther from the point of tangency.

SECTION 3

Visual Explanation — Tangent Line Approximation

The following diagram illustrates the geometric foundation of marginal approximation. A nonlinear cost function C(q) is shown as a curve, and the tangent line at a specific production level q₀ provides the linear estimate. The vertical gap between the tangent line's prediction and the actual curve represents the approximation error. Notice how for a small change Δq the tangent line tracks the curve closely, but for a larger change the error grows substantially.

The violet curve represents the actual cost function C(q). The dashed cyan line is the tangent at q₀. For a small Δq, the tangent-line estimate (cyan) nearly matches the actual change (green), producing minimal error. For a large Δq, the error grows significantly because the curve bends away from the tangent line.

This diagram encapsulates the entire logic of marginal approximation. The tangent line at q₀ has slope C′(q₀), which is the marginal cost at that production level. Moving rightward by Δq along the tangent line yields an estimated cost change of C′(q₀) × Δq (the cyan segment), while the actual cost change C(q₀ + Δq) − C(q₀) is the green segment. The red region between them is the approximation error. As Δq shrinks toward zero, the tangent line converges to the curve, and the error vanishes — this is precisely the limiting behavior that defines the derivative.

SECTION 4

Mathematical Framework

The mathematical foundation for marginal approximation rests on the tangent-line approximation (also called the linear approximation or differential approximation). Given a differentiable function f(q), the first-order Taylor expansion about a point q₀ provides the fundamental formula. We derive the key relationships below and then specialize them for cost, revenue, and profit.

GENERAL LINEAR APPROXIMATION

f(q₀ + Δq) ≈ f(q₀) + f′(q₀) · Δq

f(q₀) = the function value at the current production level; f′(q₀) = the derivative (marginal function) evaluated at q₀; Δq = the change in quantity (often Δq = 1 in business contexts).

CHANGE IN FUNCTION VALUE

Δf ≈ f′(q₀) · Δq

This isolates the estimated change: Δf = f(q₀ + Δq) − f(q₀). When Δq = 1, the marginal value directly approximates the change from producing one additional unit.

MARGINAL COST APPROXIMATION

C(q₀ + 1) − C(q₀) ≈ C′(q₀)

The cost of the (q₀ + 1)th unit is approximately equal to the marginal cost function evaluated at q₀. This is the single most commonly used marginal approximation in business calculus.

MARGINAL PROFIT APPROXIMATION

P(q₀ + 1) − P(q₀) ≈ P′(q₀) = R′(q₀) − C′(q₀)

Since P(q) = R(q) − C(q), marginal profit equals marginal revenue minus marginal cost. If R′(q₀) > C′(q₀), producing one more unit increases profit; if R′(q₀) < C′(q₀), it decreases profit.

💡

Why Δq = 1?

In most business calculus problems, we set Δq = 1 because the natural question is about producing one more unit. However, the approximation formula works for any small Δq. For instance, if you want to estimate the cost change from producing 5 more units, use ΔC ≈ C′(q₀) × 5. The approximation remains valid as long as the cost curve doesn't bend too sharply over that interval.

SECTION 5

Exact vs. Approximate — A Quantitative Comparison

To build intuition for how good (or how limited) the marginal approximation is, consider a concrete cost function: C(q) = 0.01q³ − 0.6q² + 15q + 200. We can compute both the exact change (by evaluating C at two points) and the marginal approximation (using C′(q₀)) at several production levels and for several sizes of Δq. The following table and diagram reveal how the approximation quality varies.

q₀	Δq	C′(q₀) · Δq (Approx.)	C(q₀ + Δq) − C(q₀) (Exact)	Error
10	1	$6.00	$6.21	$0.21
10	5	$30.00	$33.75	$3.75
20	1	$3.00	$3.41	$0.41
20	5	$15.00	$21.75	$6.75
30	1	$12.00	$12.61	$0.61

Comparison of marginal approximation vs. exact change for C(q) = 0.01q³ − 0.6q² + 15q + 200

Approximation error grows nonlinearly as Δq increases. Each curve shows the error at a different base production level. The pink curve (q₀ = 30) shows the largest errors because the cost function curves more steeply at higher production levels. The message is clear: keep Δq small for reliable approximations.

The data confirm two critical observations. First, when Δq = 1 the approximation error is small — typically within a few percent of the true change — making it a practical tool for unit-by-unit decision-making. Second, the error accelerates as Δq grows because the second derivative (the curvature of the function) causes the tangent line to diverge from the curve. In regions where the function has high curvature — that is, where |C″(q)| is large — even moderate values of Δq can yield poor approximations. This is why the second derivative provides a rough gauge of approximation quality.

SECTION 6

Worked Example — Marginal Cost & Revenue Approximation

A small electronics company manufactures wireless earbuds. Its total cost and total revenue functions (in dollars) are given by:

COST FUNCTION

C(q) = 0.005q³ − 0.5q² + 28q + 1500

where q is the number of units produced per day.

REVENUE FUNCTION

R(q) = 75q − 0.15q²

where q is the number of units sold per day (assume all produced units are sold).

The company currently produces q₀ = 40 units per day. Management wants to know: (a) approximately how much will total cost increase if production rises to 41 units, (b) approximately how much additional revenue will the 41st unit generate, and (c) should the company produce the 41st unit?

Marginal Cost and Revenue at q₀ = 40

1
Step 1 — Find the Marginal Cost Function

Differentiate C(q) = 0.005q³ − 0.5q² + 28q + 1500 to obtain the marginal cost function: C′(q) = 0.015q² − q + 28.

C′(q) = 0.015q² − q + 28
2
Step 2 — Evaluate Marginal Cost at q₀ = 40

Substitute q = 40: C′(40) = 0.015(40)² − 40 + 28 = 0.015(1600) − 40 + 28 = 24 − 40 + 28 = 12. Thus the estimated additional cost of the 41st unit is $12.00.

C′(40) = $12.00 ≈ cost of the 41st unit
3
Step 3 — Find the Marginal Revenue Function

Differentiate R(q) = 75q − 0.15q² to obtain: R′(q) = 75 − 0.30q.

R′(q) = 75 − 0.30q
4
Step 4 — Evaluate Marginal Revenue at q₀ = 40

Substitute q = 40: R′(40) = 75 − 0.30(40) = 75 − 12 = 63. The 41st unit is estimated to generate $63.00 in additional revenue.

R′(40) = $63.00 ≈ revenue from the 41st unit
5
Step 5 — Compute Marginal Profit and Decide

Marginal profit = R′(40) − C′(40) = 63 − 12 = 51. Since the marginal profit is positive, the 41st unit adds approximately $51 to total profit. The company should produce it.

P′(40) = $51.00 > 0 → Produce the 41st unit ✓
6
Step 6 — Verify with Exact Calculation

For comparison: C(41) − C(40) = [0.005(41)³ − 0.5(41)² + 28(41) + 1500] − [0.005(40)³ − 0.5(40)² + 28(40) + 1500] = 2336.405 − 2324 = $12.405. The exact cost change is $12.41, and our marginal approximation of $12.00 is off by only $0.41 — an error of about 3.3%. Similarly, R(41) − R(40) = [75(41) − 0.15(41)²] − [75(40) − 0.15(40)²] = 2822.85 − 2760 = $62.85. Our estimate of $63.00 differs by only $0.15.

Approximation errors: ΔC error ≈ $0.41 (3.3%), ΔR error ≈ $0.15 (0.2%)

SECTION 7

Strengths, Limitations & When to Use Marginals

Marginal approximation is an extraordinarily useful tool, but like any approximation it has boundaries. Understanding when to trust it — and when to compute exact values — is as important as knowing how to use it. The following table summarizes the trade-offs.

Strengths	Limitations
Fast computation: only requires evaluating the derivative at one point, not the entire function at two points.	Accuracy degrades for large Δq, especially when the function has high curvature (large \|f″(q)\|).
Gives immediate decision insight: if marginal profit > 0, produce more; if < 0, cut back.	Only provides the change estimate, not the absolute value of cost/revenue at the new quantity.
Works for any differentiable business function — cost, revenue, profit, or even demand.	Assumes the function is smooth and differentiable; real-world cost functions may have jumps (e.g., new equipment at capacity thresholds).
Excellent for unit-by-unit analysis (Δq = 1), the most common business scenario.	Ignores fixed-cost step changes, discrete pricing, and integer constraints on output.

Trade-offs of using marginal approximation in business analysis

⚠️

WHEN TO TRUST THE APPROXIMATION

A useful rule of thumb: if the percentage change in quantity (Δq / q₀) is less than about 5%, the marginal approximation is generally reliable for smooth business functions. Beyond that threshold, especially for cubic or higher-order cost curves, consider computing exact values. Think of it like a GPS recalculating — for small detours, the estimated arrival time barely changes, but a major reroute requires a full recalculation.

SECTION 8

Connection to Optimization & Higher-Order Approximations

The marginal approximation studied in this lesson is a first-order (linear) technique. It naturally connects to several more advanced ideas that you will encounter in deeper calculus and economics courses. The most immediate extension is profit maximization: setting P′(q) = 0 (equivalently, R′(q) = C′(q)) identifies the quantity where producing one more unit neither adds to nor subtracts from profit — the optimal production level. Beyond that, second-order Taylor approximations use the second derivative to improve accuracy for larger Δq, and elasticity analysis reframes marginal revenue in terms of percentage changes in price and quantity.

Concept	This Lesson (First-Order)	Advanced Extension
Approximation	Δf ≈ f′(q₀) · Δq (linear/tangent line)	Δf ≈ f′(q₀) · Δq + ½ f″(q₀) · (Δq)² (quadratic/Taylor)
Decision Rule	If P′(q) > 0, increase production	Set P′(q) = 0 and verify P″(q) < 0 for maximum profit
Error Analysis	Error ≈ ½ \|f″(c)\| · (Δq)² for some c between q₀ and q₀ + Δq	Remainder term from Taylor's theorem provides rigorous error bounds
Scope	One function, one variable, small Δq	Multivariate marginal analysis using partial derivatives (e.g., cost depends on quantities of multiple products)

From first-order marginal analysis to advanced optimization

Understanding the first-order marginal approximation thoroughly prepares you for these extensions. The fundamental intuition — that the derivative captures the rate of change at a point, and that this rate predicts nearby behavior — remains the core principle whether you are doing linear approximation, optimization, or multivariable analysis. The leap from 'approximate the change' to 'find where the change is zero' is precisely the leap from this lesson to profit maximization.

SECTION 9

Practice Problems

Question 1: CONCEPTUAL

Explain in your own words why C′(q₀) approximates C(q₀ + 1) − C(q₀) rather than equaling it exactly. Under what condition would the approximation become exact?

Show answer

C′(q₀) is the slope of the tangent line to the cost curve at q₀, which represents the instantaneous rate of change. The actual cost change C(q₀ + 1) − C(q₀) corresponds to a secant line over a finite interval of width 1, which captures the curvature of the function. The two values differ because the cost curve is generally nonlinear — it bends between q₀ and q₀ + 1. The approximation would be exact if C(q) were a linear function (i.e., C″(q) = 0 everywhere), because a tangent line to a straight line is the line itself. In practical terms, the flatter the curvature in the neighborhood of q₀, the better the approximation.

Question 2: BASIC CALCULATION

A company's total cost function is C(q) = 2q² + 50q + 800. Use the marginal cost to approximate the cost of producing the 26th unit (i.e., increasing production from 25 to 26 units).

Show answer

First, find the marginal cost function: C′(q) = 4q + 50. Evaluate at q = 25: C′(25) = 4(25) + 50 = 100 + 50 = 150. So the approximate cost of the 26th unit is $150. To verify, the exact cost change is C(26) − C(25) = [2(676) + 50(26) + 800] − [2(625) + 50(25) + 800] = [1352 + 1300 + 800] − [1250 + 1250 + 800] = 3452 − 3300 = $152. The approximation of $150 is off by only $2, which is within 1.3% of the exact value — illustrating the reliability of marginal approximation for Δq = 1.

Question 3: INTERMEDIATE

A bakery's daily revenue is modeled by R(q) = 120q − 0.4q², and its cost is C(q) = 0.2q² + 20q + 500, where q is dozens of croissants. The bakery currently produces 80 dozen. (a) Use marginals to approximate the change in profit from producing 3 additional dozen. (b) Should the bakery increase production?

Show answer

First, find P(q) = R(q) − C(q) = 120q − 0.4q² − 0.2q² − 20q − 500 = −0.6q² + 100q − 500. Then P′(q) = −1.2q + 100. Evaluate at q = 80: P′(80) = −1.2(80) + 100 = −96 + 100 = 4. With Δq = 3, the approximate change in profit is ΔP ≈ P′(80) × 3 = 4 × 3 = $12. Since the estimated change is positive, the bakery gains approximately $12 in profit from producing 3 more dozen, so yes, it should increase production. However, note that P′(q) is decreasing (P″(q) = −1.2 < 0), so each successive dozen contributes less marginal profit than the last, and the approximation slightly overstates the benefit for Δq = 3.

Question 4: APPLIED

A software company sells annual licenses. Its revenue function is R(q) = 500q − 0.02q³ and its cost function is C(q) = 0.01q³ + 10q + 20000, where q is the number of licenses. At q = 100 licenses, (a) compute the marginal profit, (b) use it to approximate the change in profit if sales increase by 2 licenses, and (c) determine the approximation error by comparing with the exact change.

Show answer

The profit function is P(q) = R(q) − C(q) = 500q − 0.02q³ − 0.01q³ − 10q − 20000 = −0.03q³ + 490q − 20000. The marginal profit is P′(q) = −0.09q² + 490. (a) At q = 100: P′(100) = −0.09(10000) + 490 = −900 + 490 = −$410. (b) For Δq = 2: ΔP ≈ P′(100) × 2 = −410 × 2 = −$820. The negative marginal profit indicates the company should not sell additional licenses at this level; doing so reduces profit. (c) Exact change: P(102) − P(100) = [−0.03(102)³ + 490(102) − 20000] − [−0.03(100)³ + 490(100) − 20000] = [−0.03(1061208) + 49980 − 20000] − [−0.03(1000000) + 49000 − 20000] = [−31836.24 + 49980 − 20000] − [−30000 + 49000 − 20000] = −1856.24 − (−1000) = −$856.24. The approximation of −$820 has an error of $36.24, about 4.2% of the exact change. The error is larger here because the cubic function has significant curvature at q = 100.

Question 5: CRITICAL THINKING

Suppose a firm's cost function satisfies C″(q) > 0 for all q > 0 (i.e., the cost curve is concave up). Prove that the marginal approximation C′(q₀) always underestimates the actual cost of the (q₀ + 1)th unit. Then discuss: does the analogous result hold for revenue functions? Why or why not?

Show answer

When C″(q) > 0, the cost function is concave up, meaning the curve lies above its tangent line at every point. The tangent line at q₀ predicts the cost at q₀ + 1 as C(q₀) + C′(q₀), but the actual cost C(q₀ + 1) is higher because the curve bends upward away from the tangent. Therefore C(q₀ + 1) − C(q₀) > C′(q₀), and the marginal cost underestimates the true cost of the next unit. Formally, by the Mean Value Theorem, C(q₀ + 1) − C(q₀) = C′(c) for some c ∈ (q₀, q₀ + 1). Since C″ > 0, C′ is increasing, so C′(c) > C′(q₀), confirming the underestimate. For revenue, the situation is typically reversed: most revenue functions have R″(q) < 0 (concave down) due to diminishing marginal returns in pricing. In that case, the tangent line lies above the curve, and R′(q₀) overestimates the true additional revenue. However, this is not universal — some revenue models may have R″(q) > 0 in certain ranges (e.g., network effects), which would flip the inequality. The key insight is that the sign of the second derivative determines the direction of the approximation bias.

SUMMARY

Lesson Summary

Summary

This lesson established that marginal analysis uses the derivative of a business function to approximate how costs, revenues, and profits change when production shifts by a small amount. The core formula — Δf ≈ f′(q₀) · Δq — is derived from the tangent-line (linear) approximation and works best when Δq is small relative to q₀. We saw that marginal cost C′(q) estimates the cost of one additional unit, marginal revenue R′(q) estimates the additional revenue earned, and marginal profit P′(q) = R′(q) − C′(q) determines whether producing one more unit adds to or subtracts from profit.

The approximation error depends on the curvature of the function (measured by the second derivative) and the magnitude of Δq. In the worked example, marginal approximations of cost and revenue at q = 40 differed from exact values by only 3.3% and 0.2% respectively, confirming the method's practical reliability for single-unit changes. Looking ahead, this first-order technique directly leads to profit optimization (setting P′(q) = 0) and higher-order Taylor approximations for larger intervals.