Insight and Segmentation
Explore, calculate, segment
Insight work starts in Data Explorer or the Query Editor, where you run SQL against DMOs and unified profile objects to validate data. Calculated Insights turn SQL into reusable metrics stored as metric objects - batch insights refresh on a schedule, streaming insights update near real time from streaming data. You need a calculated insight whenever a segment criterion is an aggregate (count of orders, total spend, days since last purchase). Row-level filters in the segment builder do not require one.
Segmentation
Segments are built on DMOs - most commonly the Unified Individual - with filters on profile attributes and on related attributes reached through modeled relationships. Segment membership is populated when the segment is published or refreshed, and counts are not live before then. Segments can be nested, combined, or used to exclude populations. A segment is the input to activation; on its own it delivers nothing to a destination system.
Traps
- Trying to filter on a SUM or COUNT in the segment builder without a calculated insight.
- Assuming membership updates instantly after a data change - refresh cadence governs it.
- Filtering on a related object that has no relationship modeled on the DMO.
- Confusing a segment's population with an activation's delivered audience.
- Building on a source DMO when the scenario clearly implies unified profiles.
Rule of thumb
Ask "row-level or aggregate?" Row-level filters belong in the segment builder. Aggregates, ratios, or derived time windows belong in a calculated insight first, and the segment then filters on the resulting metric.