What the compound annual growth rate tells you
CAGR is the single constant annual rate that would take a beginning value to an ending value over a stated number of years. It is a smoothing device. Nothing in the real series has to have grown at that rate in any particular year — the rate is chosen precisely so that the wobble in between disappears and only the two endpoints survive.
That is what makes it useful and what makes it dangerous. It is useful because raw cumulative growth is not comparable across different holding periods: a 60% gain means something very different over three years than over twelve. Annualising puts every investment, revenue line or user count on the same per-year footing. It is dangerous because the smoothing hides everything about the journey, including how much risk you were carrying and how the value behaved when you might have needed it.
The measure applies to anything positive and measured at two dates. Analysts use it on revenue, unit shipments, subscriber counts, dividends per share and installed capacity as readily as on portfolio values. What it cannot handle is a quantity that crossed zero, or one where money moved in and out during the period — those need different tools, covered below.
Why the formula has the shape it has
Start from the definition of compounding. If a value grows at a constant rate g for n years, then the ending value equals the beginning value times (1 + g)n. You know the two values and the time; you want the rate. So divide both sides by the beginning value, take the nth root of the ratio, and subtract one. That is the whole derivation: CAGR = (Vf / Vi)^(1/n) − 1.
Three consequences follow directly from that algebra. First, only the endpoints matter — substitute any path you like between them and the CAGR is unchanged. Second, the rate is undefined unless both values are positive: a negative or zero ending value has no real nth root of the ratio in the relevant sense, and no positive base raised to a power ever reaches zero. Third, the exponent 1/n means that measurement error in the dates is amplified when n is small. Get the period wrong by a month on a 20-year series and the rate barely moves; get it wrong by a month on a six-month series and it moves a great deal.
The doubling time falls out of the same relation. Set the ending value to twice the beginning value and solve for n: 2 = (1 + g)n, so n = ln 2 ÷ ln(1 + g). This is exact. The familiar rule of 72 is a linear approximation to it and drifts as rates rise, which is why the calculator reports the exact value.
The comparison output is deliberately a subtraction in percentage points, not a ratio. Comparing two rates by dividing one by the other produces a number nobody can interpret; the gap in percentage points is what an investment committee actually discusses.
Worked example: $10,000 becomes $20,000 over ten years
Run the default inputs through by hand.
- Growth ratio. $20,000 ÷ $10,000 = 2.000000.
- Exponent. 1 ÷ 10 = 0.1.
- Annual growth factor. 20.1 = 1.07177346. Check it by squaring twice and multiplying: 1.071773462 = 1.14869835, 1.148698352 = 1.31950791, and 1.319507912 = 1.74110113, which is 20.8; multiply by 1.14869835 to reach 2.000000.
- CAGR. 1.07177346 − 1 = 0.07177346, or 7.1773% a year.
- Total growth. 2.000000 − 1 = 100% over the whole ten years.
- Doubling time. ln 2 ÷ ln 1.07177346 = 0.693147 ÷ 0.0693147 = 10.00 years, which must be true by construction because the value did in fact double in ten years.
- Gap to a 7% comparison rate. 7.1773 − 7.0000 = 0.177 percentage points.
- Projected five years further. $20,000 × 1.071773465 = $20,000 × 1.414214 = $28,284.27. Five years is half a doubling period, so the factor is the square root of two — a useful check.
Every intermediate figure here is reproducible on any calculator with a power key, and the last two steps are exact by construction rather than coincidence.
How to read the number you get
Judge a CAGR against three things: an alternative, a horizon, and the risk taken to earn it.
Against an alternative. A growth rate in isolation is meaningless. The comparison-rate input exists so that you always produce a gap rather than a bare number. Use the index you would otherwise have bought, the hurdle rate your board applies, or your cost of capital — which the WACC calculator will give you if you are assessing a business rather than a portfolio.
Against the horizon. Short-period CAGRs are noise amplifiers. A 4% gain over one quarter annualises to 17%, and reporting that as an annual rate implies a persistence the data cannot support. The Global Investment Performance Standards, in their 2020 edition, require that returns for periods of less than one year must not be annualised in reported performance. The calculator will still compute it, because you may have a legitimate internal reason to, but it warns you on any holding period under a year and tells you the exponent your observation is being raised to.
Against the risk. Two investments with identical CAGRs can have utterly different paths. The smoothed rate says nothing about the largest drawdown along the way, and a path that spent time deeply underwater is a different product from one that rose steadily, even when the endpoints match.
Finally, be careful with the doubling-time output when the rate is small. At 1% a year a doubling takes about 70 years, which is a correct answer to an unimportant question. Doubling time is most informative in the 5–20% band, where the difference between 14 years and 4 years is a difference you can act on.
CAGR for a given total growth over a given period
| Ending value as a multiple of the start | 3 years | 5 years | 10 years | 20 years |
|---|---|---|---|---|
| 1.25× (+25%) | 7.722% | 4.564% | 2.257% | 1.122% |
| 1.5× (+50%) | 14.471% | 8.447% | 4.138% | 2.048% |
| 2× (+100%) | 25.992% | 14.870% | 7.177% | 3.526% |
| 3× (+200%) | 44.225% | 24.573% | 11.612% | 5.647% |
| 5× (+400%) | 70.998% | 37.973% | 17.462% | 8.380% |
| 10× (+900%) | 115.443% | 58.489% | 25.893% | 12.202% |
Every cell is multiple^(1/years) − 1. Note how flat the columns become as the horizon lengthens: a tenfold gain over 20 years is a lower annual rate than a mere doubling over 3.
Where CAGR misleads
- Cash flows during the period. If you added $5,000 halfway through, the ending value reflects your deposit as well as growth, and the CAGR overstates performance. Use a money-weighted return — the internal rate of return calculator handles dated cash flows properly.
- Cherry-picked endpoints. Because only two dates enter the formula, moving the start date by a few months across a market bottom or top can change the rate dramatically. Always state the exact endpoints alongside the rate, and be suspicious of any reported CAGR that does not.
- Treating it as an expected return. A historical CAGR is a description of what happened between two dates. Projecting it forward, which the projection output lets you do, is an assumption you are making, not a result the data supports.
- Comparing it with an arithmetic average return. The average of yearly returns is always at least the CAGR, and the gap widens with volatility. A series of +30% then −20% averages +5% but compounds at 1.98% a year, because 1.30 × 0.80 = 1.04 and 1.04^0.5 = 1.0198.
- Applying it to a quantity that can be negative. Operating profit, free cash flow and net income cross zero routinely. A CAGR on a series that starts negative is meaningless, and one that ends negative is undefined. Report the absolute change instead.
- Ignoring inflation. A nominal CAGR of 6% during a period of 4% inflation is a real rate of about 1.92%, from 1.06 ÷ 1.04 − 1. Decide which one you mean before you quote it.
Do not annualise short periods in reported performance
The GIPS standards for firms, 2020 edition, state that returns for periods of less than one year must not be annualised. The reason is not fussiness: annualising a two-month result raises it to the sixth power, and any noise in the measurement is raised to the sixth power with it. A 3% move over two months becomes 19.4% annualised, which reads as a claim about a year that two months of data cannot support. Present short periods as the cumulative return for that period, clearly labelled.
When to reach for a different measure
CAGR answers one narrow question: what constant rate connects these two numbers. Several neighbouring questions need different machinery.
If money moved in or out, the honest measure is money-weighted. The internal rate of return solves for the discount rate that sets the net present value of all dated cash flows to zero, which is the generalisation of CAGR to an arbitrary cash flow stream — with a single outflow at the start and a single inflow at the end, the IRR and the CAGR are the same number.
If you want to know what a rate produces going forward, you are compounding rather than measuring. The compound interest calculator handles a starting balance plus regular deposits, and the future value calculator works in periods and a periodic rate for coursework-style problems.
If you want today's value of a future amount, run the same relation backwards with the present value calculator. Discounting at rate g for n years is exactly compounding at rate g for −n years.
If you are appraising a project rather than measuring an asset, the growth rate of anything is the wrong headline number. What matters is whether the flows, discounted at your cost of capital, exceed the outlay — the net present value calculator answers that directly.
Used inside those limits, CAGR is the cleanest one-number summary of growth there is. It has exactly three inputs, no conventions to argue about, and an unambiguous definition — which is more than can be said for most performance statistics.