New launch, old phone

Every fall, when a new flagship phone is announced, searches for it spike. And anyone who already owns a phone asks the practical question: what is mine still worth? Should I trade it in now or hold on another year?

The answer runs on math you’ve seen in high school: exponential functions and percentages. All prices here are hypothetical, picked so the arithmetic stays clean. The goal isn’t to predict the resale market; it’s to understand the model behind any trade-in or resale value chart.

The model: losing a fixed percentage each period

Say a device loses 30% of its value each year. The key question is: 30% of what? In year one, 30% of the purchase price. In year two, 30% of whatever is left, not of the original price. Each year multiplies the value by 0.7.

Losing r percent is the same as multiplying by (1−r)(1 - r). Doing that t times gives a power:

V(t)=V0⋅(1−r)tV(t) = V_0 \cdot (1 - r)^t
V₀ is the starting value, r the loss rate per period (as a decimal), and t the number of periods.

Because the base 0.70.7 is between 0 and 1, the function is decreasing: the value keeps dropping but never hits zero. The dollar drops get smaller each year, since the percentage applies to a shrinking amount. That’s classic exponential decay, the same behavior as a drug clearing your bloodstream or a radioactive sample.

$1200Year 0$840Year 1$588Year 2$411.6Year 3
Each year multiplies the value by 0.7: $1,200 → $840 → $588 → $411.60. The dollar drops shrink because 30% applies to a smaller amount.

Compared with straight-line depreciation

There’s another model, common in accounting and on tax forms: straight-line depreciation. The item loses the same dollar amount each year until it reaches zero after n years. Depreciating the 1,200phoneover5yearswouldmeanlosing1,200 phone over 5 years would mean losing240 per year.

L(t)=V0−V0n tL(t) = V_0 - \frac{V_0}{n}\, t
Straight-line depreciation: a constant loss of V₀/n per period, reaching zero after n periods.

Worked example: the $1,200 phone

Year-by-year value at a 30% yearly loss
  1. 1−0.30=0.7\displaystyle 1 - 0.30 = 0.7 Write the rate as a decimal and find the yearly factor: 30% = 0.30, so the factor is 1 − 0.30 = 0.7.
  2. V(t)=1200⋅0.7t\displaystyle V(t) = 1200 \cdot 0.7^t Set up the function with the hypothetical $1,200 price.
  3. V(1)=840\displaystyle V(1) = 840 After 1 year: 1,200 × 0.7 = $840.
  4. V(2)=1200⋅0.49=588\displaystyle V(2) = 1200 \cdot 0.49 = 588 After 2 years: 1,200 × 0.49 = $588.
  5. V(3)=1200⋅0.343=411.6\displaystyle V(3) = 1200 \cdot 0.343 = 411.6 After 3 years: 1,200 × 0.343 = $411.60.
  6. L(3)=1200−720=480\displaystyle L(3) = 1200 - 720 = 480 Compare with 5-year straight-line: after 3 years, L(3) = 1,200 − 3 × 240 = $480.

Check. Go year by year: 1,200 − 30% = 840; 840 − 30% (252) = 588; 588 − 30% (176.40) = 411.60. Matches the power. The total loss over 3 years is 1,200 − 411.60 = 788.40, and 788.40 / 1,200 is 65.7%, not 90%.

Notice the comparison too: early on, the exponential model drops faster than the straight line (840vs.840 vs.960 after year 1). But in year 5 the straight line hits zero, while the exponential still says $201.68. Which model fits a real phone depends on real data; what matters is knowing they make very different predictions.

012345020040060080010001200yearsvalue ($)Exponential (30%/yr)Straight-line (5 yrs)
The exponential drops faster at first ($840 vs. $960 after year 1), but the straight line hits zero in year 5, when the exponential still says $201.68.

Half-life: how long until the value is cut in half

“When is the phone worth half?” asks for the t that solves 0.7t=0.50.7^t = 0.5. The unknown is in the exponent, so the tool is the logarithm:

t1/2=ln⁡(1/2)ln⁡(1−r)=ln⁡0.5ln⁡0.7≈1.94t_{1/2} = \frac{\ln(1/2)}{\ln(1 - r)} = \frac{\ln 0.5}{\ln 0.7} \approx 1.94
Half-life at a 30% yearly loss; it holds for any starting value.

That’s about 1.94 years, just under 2. Sanity check: 0.72=0.490.7^2 = 0.49, slightly under half, as expected. And the half-life doesn’t depend on the starting price: a 1,200phoneanda1,200 phone and a500 phone take the same time to lose half their value if the rate is the same. By the same logic, dropping to a quarter takes two half-lives, about 3.89 years.

49%49 / 1000.7² = 0.49 of thestarting value
After 2 years, 49% of the value is left (0.7² = 0.49), just under half: that's why the half-life is about 1.94 years.

Backing out the rate from a trade-in offer

In real life you usually have the reverse: you know what you paid, and you get a trade-in quote. Say the phone cost 1,200andtwoyearslatertheofferis1,200 and two years later the offer is588. What yearly rate does that imply?

Solve for r: (1−r)2=588/1200=0.49(1 - r)^2 = 588/1200 = 0.49, so 1−r=0.71 - r = 0.7 and r = 30%. In general:

r=1−(VtV0)1/tr = 1 - \left(\frac{V_t}{V_0}\right)^{1/t}

Common mistakes

  • Adding the percentages. 30% a year for 3 years isn’t 90%. Successive percentages multiply: the factor is 0.73=0.3430.7^3 = 0.343.
  • Always applying the rate to the original price. That quietly turns the model into straight-line.
  • Dividing the total loss by the number of years. Losing 51% in 2 years is not 25.5% a year; it’s 30% a year.
  • Expecting the value to hit zero. In the exponential model it only approaches zero.
  • Mixing up the rate and the factor. The rate is 0.3; the multiplier is 0.7.

Frequently asked questions

Why model phone depreciation as exponential rather than linear?

It's a model, not a law. When the loss is proportional to current value (a percentage), you get exponential decay. Real prices can drop sharply at a new launch; exponential is a useful approximation.

How do I estimate the half-life without a scientific calculator?

Try powers: 0.7² = 0.49, so the half-life is just under 2 years. For precision, use t = ln 0.5 / ln(1 − r).

What if the loss is monthly?

Same formula, with t in months and r the monthly rate. A 2% monthly loss gives a yearly factor of 0.98¹² ≈ 0.785, roughly a 21.5% yearly loss.

Is trade-in value the same as resale value?

They're two different prices for the same phone. The math is the same: with two data points (price paid and offer) you can compute the implied rate and compare offers.