Audiences built on the operator's own data reach the right player 1.4x more often than the network's native audiences
Networks report impressions and post-view deposits, but not whether an impression reached the player you paid for. We measured it directly, on 4.5 million impressions a day, for two operators in South Asia.
Situation
Retargeting in Ubidex ran on two kinds of audiences: the network's native retargeting audiences (its pixel and its matching) and the operator's own segments built from CRM and tracking data and sent to the network server-side. Which of them actually reaches the targeted player was unknown.
What we did
- Joined the network's impression log with the operator's visit log by IP and user agent, every day, with a completeness check against the network dashboard (1.00 every day of the period).
- Labeled an impression a hit when the targeted player (the user id the network was asked to reach) came back to the site within one hour.
- Compared hit rate by audience type, by week and by ad format, with 95% confidence intervals, and set it up as a daily job with a dashboard per source x campaign.
Result (28 September to 3 November 2026)
| Audience type | Operator A, hit rate | Operator B, hit rate |
|---|---|---|
| Native network audiences | 0.256 (n 118,816) | 0.298 (n 166,764) |
| Operator's own segments | 0.355 (n 2,952) | 0.434 (n 3,231) |
| Ratio | 1.38x | 1.45x |
Own segments led in every one of the six weeks and within each format: pop 0.382 vs 0.224 and push 0.321 vs 0.221 for operator A; push 0.427 vs 0.238 for operator B. Native sources differ from each other by more than 10x in precision (from 0.05 to 0.67), which is exactly the information a buyer needs to decide where to bid.
What to keep in mind
- Hit rate is a proxy for reaching the right player, not a deposit metric; we did not find a pair-level link between precision and cost per deposit in this period.
- IP plus user agent produces some false joins on mobile networks: 1.5 to 3.8% of labeled pairs carry more than one user id, which lowers absolute precision by at most 2 to 3 points and does not change the ratio.
- Operator A's impressions fell about 7x over the period as low-precision sources were switched off, so part of the rise in native precision for A is the optimization itself.
What it means for an operator
You see which share of the budget reaches the player you are paying for, and which audiences and sources deserve the bid. Segments built on the operator's own CRM data beat the network's matching, so retargeting can be built on your data instead of the network's black box.
Check your targeting: 6-week pilot