Reading a compound annual growth rate

Because the rate is built from only two observations, a steady climb and a path that triples, collapses and recovers can share exactly the same figure.

Both series compound at 14.9 percent a year. One grew at a constant rate. The other tripled, fell below its starting point, and climbed back. A reading that rests on the rate alone treats those as the same company.

The start year is the whole answer

Both rates are correctly computed from the same six filed observations. Starting one year later, in the year revenue fell from 100 to 80, nearly doubles the published growth rate. Checking what the base year looked like takes one minute and changes the reading more than any other check on this page. A third, common error: six observations span five years, not six, and using six as the exponent understates the rate to 8.1 percent. Count the gaps, not the rows.

Where the arithmetic stops meaning anything

A negative starting value gives a negative ratio, and the rate either fails to compute or computes to nothing meaningful. Where both figures are negative, a widening loss prints as growth and a narrowing one as decline. A period containing an acquisition measures a standalone company at one end and a combined one at the other, saying nothing about whether either grew. And a period crossing a company's adoption of ASC 606 can be dominated by a change in whether revenue is reported gross, as principal, or net, as agent, rather than by any change in the business.

Per share against total, which are different claims

A per-share growth rate contains the change in the share count: buybacks raise it, issuance lowers it, with no change in the business. Earnings per share uses a weighted average share count, so a large repurchase affects the following year in full but the completion year only in part. A per-share rate that accelerates a year after a buyback programme is showing the share count rather than the earnings. Putting the earnings-per-share rate next to the net-income rate separates the two.

The honest presentation of a rate

Print the path; every filed year is public and showing it costs no more than showing the summary. Where a single number is wanted, the worst single year in the period and a stability measure both do more work than the headline rate alone. ValueScreener computes revenue and earnings compound rates over three and five years alongside stability measures, and withholds a rate for which too few filed years exist rather than computing it from a shorter series.

Check yourself