Average percentage change

Each interval percent is a valid story on its own. The average of those percents is another story: "if I treat every step equally, what is the central tendency of relative moves?" That may diverge from the percent you would get by comparing only the first and last level because compounding is nonlinear.

Weighting by revenue or volume is often fairer than a naive mean when segments differ in size.

Step one remains ((N − O) ÷ |O|) × 100% for each consecutive pair. Summarize with care: arithmetic mean, median, weighted mean, or geometric interpretations each answer different questions.

When endpoints matter more than intermediate path, CAGR may replace a mean of monthly percents. Compare both and disclose which view you show externally.

1. List periods in chronological order.
2. Compute percent change for each hop using consistent baselines.
3. Inspect for outliers driven by tiny denominators.
4. Choose a summary statistic that matches the decision you support.
5. Cross-check against first-to-last percent change for intuition.

Three months of user counts produce monthly percents of +10%, −5%, and +4%. The arithmetic mean is about +3%, but the compounded path differs from comparing month one to month four directly. Tell readers which view you charted.

Link forward to CAGR when executives ask for a single headline number across years.