demo · v134

Dedup on vs off: the overcounting problem

The whole point of reading interest groups in a Shared Storage worklet is to avoid double-counting users who qualify for multiple campaigns. Without deduplication the reach histogram overcounts; with deduplication each user is counted once. Toggle between the two modes and watch the histogram change.

Shared Storage API: checking…

Simulated audience — interest group memberships:

Unique users
Raw count (no dedup)
Deduped count (v134)
Overcount factor

Per-campaign reach count:

Simulated worklet logic


  

why this matters

Without interestGroups(), a Shared Storage worklet counts every qualifying impression event once per campaign. A user who was shown ads for three campaigns gets counted three times in the aggregate reach report. With v134’s interestGroups() call, the worklet can identify overlapping group memberships and emit a single contribution per user.

see also