In practice, the distinction becomes clear when you test aggregation across dimensions, especially time.

Additive measures can be summed across all dimensions.
Example: revenue, sales quantity. Summing across products, regions, or months still makes sense.

Semi-additive measures can be summed across some dimensions, but not all, typically not across time.
Example: account balance or inventory levels. You can sum across branches, but summing month-end balances across time is misleading.

Non-additive measures cannot be meaningfully summed across dimensions.
Example: ratios, percentages, averages like profit margin or conversion rate. These require weighted calculations instead of simple sums.

A practical rule:
If summing changes the business meaning, it is not fully additive. Always validate aggregation logic against the real-world interpretation of the metric.