Demographic segmentation was useful when individual purchase data was hard to get. It is still the default in most marketing tools, which is why most marketing messages feel generic.
Behavior is a better predictor. A customer who buys every three weeks tells you when the next order is likely, what category they buy, and what to offer them. A demographic label like 'women 25–34, urban' tells you almost nothing actionable about when to email them or what to put in the subject line.
The behaviors that matter most are recency, frequency, and value — how long since the last purchase, how often a customer buys, and how much they spend. Together they form a picture of where a customer sits in the lifecycle and what message will land.
Behavioral segments are also self-updating. A customer who was 'loyal' three months ago can become 'at-risk' today without filling out a survey. The data updates the segment, and the segment updates the campaign. That is the difference between a segment you set once and a segment that actually keeps working.