Overlapping Segments Inflating Frequency
A travel brand runs three segments: browsers (visited any destination page in 30 days), searchers (used the search tool in 30 days), and bookers (completed a booking in 90 days). All three feed the same display destination. A media agency reports that users are being hit 6–8 times per day, far above the 3-impression cap. The segments were built independently by different teams. There is no suppression segment. Reporting shows each segment has healthy unique-user counts, so nobody flagged a problem.
1. Why do healthy per-segment counts fail to reveal the frequency problem, and what would you examine?
Read the worked answer
Per-segment unique counts are marginal measures - they count each segment in isolation. Frequency problems come from overlap (the same user qualifying for multiple segments, each firing the destination) and from lack of suppression. A user who browses, searches, and books can be in all three, so three destination calls stack.
What to examine:
- Segment overlap / audience overlap analysis. Quantify how many users are in 2 or 3 of the segments. High overlap explains stacked delivery.
- Destination configuration. Confirm whether each segment maps to a separate destination call or a shared one, and whether any deduplication exists at the destination.
- Recency windows.
bookershas a 90-day window, so recent buyers keep getting prospecting ads - a classic frequency culprit. - Suppression logic. There is none, which is the structural gap.
The lesson: judge audience health by cross-segment overlap and downstream delivery, not by each segment's standalone count.
2. Propose a structural fix that reduces frequency without gutting reach.
Read the worked answer
Restructure around mutually exclusive priority tiers plus suppression, rather than three parallel segments:
- Create a
recent_bookerssuppression (e.g., booked in last 90 days) and exclude it frombrowsersandsearchers. Buyers should not be prospected. - Make
searchersandbrowsersnon-overlapping by excluding searchers from the browser segment, or by giving searchers higher priority. This ensures each user triggers one prospecting call, not two. - Keep the destination cap as a backstop but do not rely on it as the primary control - caps at the destination are the last line of defense, not a substitute for clean segment logic.
- Validate with overlap analysis after the change to confirm the stacked population collapses.
This preserves reach (everyone still qualifies for exactly one relevant tier) while removing the multiplicative delivery. Re-measure frequency after a full cycle before declaring success.