First-Order Approximations in Your Head

Most mental-math shortcuts — estimating square roots, doubling times, small percentage compounding — are really the same idea wearing different clothes: near a point you already know the answer for, a smooth function looks almost like a straight line, and a straight line is easy to evaluate. This is the mental-math version of a first-order Taylor approximation, and once you see the pattern once, you start recognizing it everywhere.

The one formula behind all of it

For a smooth function $f$ and a small step $\Delta x$ away from a known point $x_0$:

$$f(x_0 + \Delta x) \approx f(x_0) + f'(x_0)\,\Delta x .$$

In plain English: start from a value you already know exactly, $f(x_0)$, and correct it by (the function's local slope) times (how far you stepped) — you're approximating the curve with its tangent line, which is accurate as long as $\Delta x$ is small relative to how fast the function is curving.

Worked example 1: 1.03^20 without a calculator

Let $f(x) = (1.03)^x$ approached differently — actually the cleanest route is via $f(r) = (1+r)^{20}$ near $r=0$: $f(0) = 1$, and $f'(r) = 20(1+r)^{19}$, so $f'(0) = 20$. Then:

$$(1.03)^{20} \approx 1 + 20 \times 0.03 = 1.6 .$$

The true value is $(1.03)^{20} \approx 1.806$ — the linear approximation undershoots noticeably here because 20 compounding periods isn't really "small," a useful reminder that first-order approximations degrade as the step or the number of compounds grows. It's best trusted for a handful of periods or genuinely small $\Delta x$, and worked example 2 shows where it shines.

Worked example 2: (1.002)^50, where it shines

Here the per-period rate is genuinely tiny. Using $f(r)=(1+r)^{50}$, $f(0)=1$, $f'(0) = 50$:

$$(1.002)^{50} \approx 1 + 50 \times 0.002 = 1.10 .$$

The true value is $(1.002)^{50} \approx 1.105$ — accurate to within half a percent, reached with a single multiplication. The lesson: first-order approximations are trustworthy when the total deviation ($n \times r$ here) stays small, not just when $r$ alone is small.

What this means in practice

This is the single unifying idea behind estimating square roots, small percentage compounding, and quick sensitivity checks on option Greeks (delta is literally a first-order approximation of an option's price change) — recognizing "I'm near a point I know, and the step is small" is worth more in an interview than memorizing a dozen unrelated tricks. The failure mode to watch for is using it when the step isn't actually small, which silently produces a wrong answer that still "feels" plausible.