One way to start testing for and discovering segments is to dive into the data we already have, whether that's in an Excel spreadsheet, within your Email Service Provider (commonly referred to as an ESP), or in an Oracle Database 12c Enterprise Edition database (commonly referred to as “OD12cEE”, or just “Oracle database”). You can start by focusing on basic information, such as age and gender, or on more complex information such as the first product customers buy.
Another, often overlooked, way to discover segments is by actually running experiments to gather information. For example, a sock retailer might want to know which of their customers likes red socks. Maybe some of their customers have already purchased red socks, and that's a good indication, but they also want to know who else might want to buy red socks. The retailer could send out an email highlighting red socks to every customer and look at who opens the email, who clicks through, and who actually makes a purchase. This type of information lets the retailer know who else is interested in the product.
It's helpful to keep a baseline in mind when running experiments so you can judge whether the segmentation is meaningful. When running experiments on segments, set aside a control group. It's much less meaningful to say that a segment had a 10% conversion rate to an email without knowing what the segment's conversion rate in the absence of that email. In addition, when comparing a segment to the entire population, you want to remove the segment from the entire population so that its performance doesn't affect the entire population's results. In other words, you want to compare segment to non-segment, not segment to non-segment plus segment.