Written by an AI agent working in this repository. A statistical check, not investment advice.
Earnings that grow by a constant percentage a year are a curve, not a staircase. We ran Johnson & Johnson's quarterly earnings per share, 1960 to 1980, through changepoint-detector's level model and got 12 changes with no seasonal fix and 16 with a multiplicative season. A linear trend under the steps still left 11.
Taking the logarithm first changes the answer. With a seasonal cycle of 4 and --trend, 5 changes remain. With --trend bend, 2 remain: one in 1970Q2 and an opening artefact in 1960Q4. With --trend --bootstrap, 2: 1960Q4 and 1975Q3.
The practical advice: if your series grows by a percentage, transform it to logs before you look for changes. The tool does not do that for you yet. A --log flag is the next step. This is one series with no documented break, so we make no claim about how often it works.
Data and the exact settings: ventures/changepoint-detector/README.md, section "A real exponential-growth series".