Campaign Delivery Audit
This sample was produced by the real Meyppi pipeline on an illustrative synthetic campaign, so every number below is a genuine engine output, including the partial detection and the version mismatch. The scenario was run with a fixed clock for reproducibility, which is why event timestamps are round; the confidence values are the classifier's policy constants. Client, campaign and publisher identifiers are redacted. A paid audit delivers this exact format with your campaign's data.
Placement detail
| Content | Publisher | Expected | Observed | Classification | Confidence |
|---|---|---|---|---|---|
| EP-1041 | Publisher A (redacted) | CR-A v1 · 30s | CR-A v1 · 30.0s @ 00:01:30 | Detected — Full | 0.99 |
| EP-1042 | Publisher A (redacted) | CR-A v1 · 30s | CR-A v1 · 24.2s (tail missing 5.75s) | Detected — Partial | 0.90 |
| EP-1043 | Publisher B (redacted) | CR-A v1 · 30s | CR-A v2 aired (version substituted) | Version Mismatch | 0.90 |
| EP-1044 | Publisher B (redacted) | CR-A v1 · 30s | No occurrence found | Not Detected | 0.95 |
| EP-1045 | Publisher C (redacted) | CR-A v2 · 30s | CR-A v2 · 30.0s | Detected — Full | 0.99 |
Rather poke at it yourself?
The interactive version of this report lets you filter by classification and open any placement to inspect its evidence: the verification event and where the ad landed in the episode.
Evidence: one verification event
This is what sits behind a single row of the table: the full verification event for EP-1041, as exported with the report. The offset, the coverage ratio, the match count, the file hash. In a paid audit you receive one of these for every placement, and the raw evidence is never discarded after classification.
{
"event_id": "VE_CMP_XXXXX_CR-A_v1_EP-1041",
"campaign_id": "CMP-XXXXX",
"creative_id": "CR-A",
"version_tag": "v1",
"source_id": "SRC_POD-1",
"content_id": "EP-1041",
"timestamp": "2026-10-07T12:00:00+00:00",
"match_type": "full",
"duration_sec": 30,
"evidence_tier": 2,
"confidence": 0.99,
"processing": {
"engine": "meyppi-fingerprint 1.0",
"attributed_version": "v1",
"content_sha256": "9f2c…a41d (redacted)"
},
"evidence": {
"offset_sec": 90.001,
"covered_time_sec": 30,
"coverage_ratio": 1,
"match_count": 3350,
"missing_head_sec": 0,
"missing_tail_sec": 0
}
} Methodology
Master creatives are fingerprinted (Wang-style constellation fingerprinting) and matched against the published media with temporal alignment. Reference coverage, continuity, gap analysis, and alignment consistency determine each classification. When the evidence does not support a call, the placement is marked Insufficient Signal or Manual Review Required, and the reason is stated in the report.
Classification definitions
- Detected — Full: the creative aired, complete delivery.
- Detected — Partial: the creative aired, delivery incomplete (e.g. trimmed duration).
- Version Mismatch: a different version of the creative aired.
- Not Detected: no matching evidence found in the published episode.
- Insufficient Signal: media quality too low for a defensible result.
- Manual Review Required: evidence ambiguous; reviewed by an analyst.
Meyppi reports algorithmic findings from acoustic fingerprint evidence. Algorithmic confidence is not legal proof.