What does it mean that anomalies are evolution signals?

When reality repeatedly contradicts your schema the schema needs updating.

Example: A product manager notices three separate user complaints about the same "edge case" in one week. Rather than patching each individually, she recognizes the pattern as a signal that her mental model of user behavior is wrong — the edge case is actually the main case for a segment she had not modeled.

Try this: Open your journal, task manager, or notes from the past two weeks. Look for three instances where reality surprised you — a prediction that missed, a conversation that went sideways, a decision that produced unexpected results. Write each on its own line. Now ask: do these point to the same underlying schema that needs revision? If two or more cluster around a single assumption, you have found your evolution signal.