Benchmarks
This page lists numbers you can check yourself. Each one is either a setting in the code or a measurement of the live system, with the date it was taken. AdPriva has not published throughput, latency or detection-accuracy benchmarks.
The tag
| What | Value | How to check |
|---|---|---|
Size of tag.js | 21,414 bytes, about 8.4 KB gzipped (measured 2026-10-23) | curl -s https://cdn.adpriva.com/tag.js | wc -c, and the same piped through gzip -c |
| When the view event is sent | About one second after the page loads | The tag’s VIEW_DWELL_MS setting |
| Requests per page load | One view event. Consent forms send their own request when submitted | Browser developer tools, Network tab |
| Retries | One retry, 1.5 seconds later, after a network error or a server error | The tag’s delivery code |
Batching and anchoring
| What | Value | How to check |
|---|---|---|
| When a batch is made | When 100 proofs are waiting, or 5 minutes after the previous batch, whichever comes first. The anchor worker checks once a minute | Anchor worker settings |
| Proofs per batch | At most 100. At current traffic, batches hold one to a few proofs | proof_count on GET https://api.adpriva.com/public/anchors/recent |
| Merkle tree depth | 20 levels, so each receipt has 20 path entries | merkle_path on GET https://api.adpriva.com/public/proofs/{proof_id} |
| Zero-knowledge proofs | None for new view batches. Batches anchored before 2026-09-28 carry a zkVerify attestation. Consent batches never had one | zkverify_status on the same endpoints |
| Gas per anchor transaction | About 28,250 gas on Horizen, whatever the batch size (measured on ten anchor transactions, 2026-10-07) | gasUsed in the receipt of any anchor_tx |
The anchor contract emits an Anchored event and stores nothing, so the cost of a transaction does not grow with the number of proofs in the batch.
What is not measured here
- Detection accuracy. Verdicts run in shadow mode. AdPriva has not published how often they are right.
- Ingest latency and throughput. No load test results are published.
- Time to anchor. A proof is batched within about 5 minutes. The Horizen transaction adds to that, and the time varies.