Just finished a 4-week A/B test on two major CDP identity graphs. Measured match rates and audience activation accuracy for a segment of 2.1M known users.
Setup:
* Traffic split 50/50 at the tag level.
* Test segment: "High-Value Cart Abandoners" (logged-in users only).
* Activation channel: Meta CAPI.
* Primary metric: attributed purchase events from the activated audience.
Raw results:
| Metric | CDP A | CDP B |
| :--- | :--- | :--- |
| Avg. Identity Resolution Rate | 78% | 82% |
| Activated Audience Size | 512k | 547k |
| Attributed Purchases | 18,432 | 22,905 |
| Effective CPM (audience cost) | $12.40 | $11.80 |
Key finding: The 4% higher match rate for CDP B directly translated to ~24% more attributed purchases, despite only a ~7% larger activated audience. Cost efficiency was better.
Config used for the test audience export:
```json
{
"segment_id": "cart_abandon_7day",
"export_destination": "meta_audiences",
"identifier_priority": ["email", "phone", "external_id"]
}
```
Graph B consistently performed better on email+phone deterministic matches. Graph A had more unidentified device graphs. For pure performance on logged-in user activation, Graph B is the clear result.
- bench_beast
Benchmarks don't lie.