Predictive segmentation refers to the capability of automatically identifying and creating meaningful visitor segments characterized by a higher probability to react in a certain manner to specific events or conditions.
Prior to the introduction of predictive segmentation, the task of analyzing, identifying, and segmenting online traffic was painstakingly manual. The process required marketers to slice analytics data and analyze each segment at a time in an effort to reveal actionable insights or common behavior related to the primary KPI (read further about behavioral segmentation). Identifying segments manually is achieved by utilizing either a bottom-up approach, in which traffic is segmented by each available data-point and then measured against the primary KPI, or alternatively, a top-down approach where traffic is segmented according to KPI performance, and then broken down per data-point.
Available data-points often include:
- Geographic data
- Browser and OS data
- Marketing channel
- New vs. returning visitor
- Demographic data
- Life Time Value
After all traffic is segmented, marketers would be required to sift through and identify the most valuable or underperforming segments, and then act upon the data by targeting each segment and providing an experience tailored to each. Best practices require each targeted experience be A/B tested, in order to guarantee the right experience is delivered to each segment.
Predictive segmentation removes all of the manual work by automatically identifying and surfacing the valuable or high-potential traffic segments that should be targeted. However, as a standalone capability, it does not aid in determining the right experience that should be tailored for each identified segment. In other words, predictive segmentation advances a marketer’s position in the workflow, providing them with the proper segments to work with, but tailoring experiences remain guesswork that ought to be tested and validated. The latter challenge is addressed by new technologies such as “Predictive Targeting” via platforms like Dynamic Yield (read further about predictive analytics).
