RELIABILITY
How much should you trust this data?
Scanning the open web is noisy. This page publishes our own error surface: scan success rates, how false positives are prevented, every manual correction we have issued, and how to dispute a data point.
Current measurements
94.5%
of per-agent checks returned a conclusive verdict in the latest scan (8,636 of 9,135)
72
brands with ≥1 inconclusive agent verdict (of 1,015 scanned)
98.2%
average per-run scan completion across the last 14 pipeline runs
Recent scan runs
| Started (UTC) | Status | Brands | Completed | Failed | Changes |
|---|---|---|---|---|---|
| 2026-06-20 06:14 | completed | 1015 | 1007 | 8 | 18 |
| 2026-06-19 06:58 | completed | 1015 | 1007 | 8 | 63 |
| 2026-06-18 06:45 | completed | 1015 | 1007 | 8 | 54 |
| 2026-06-17 07:01 | completed | 1015 | 1008 | 7 | 9 |
| 2026-06-16 07:19 | completed | 1015 | 1007 | 8 | 27 |
| 2026-06-15 07:10 | completed | 1015 | 1007 | 8 | 0 |
| 2026-06-14 06:28 | completed | 1015 | 1007 | 8 | 270 |
| 2026-06-13 06:10 | completed | 1015 | 1015 | 0 | 27 |
| 2026-06-12 06:27 | completed | 1015 | 1015 | 0 | 27 |
| 2026-06-11 06:42 | completed | 1015 | 1015 | 0 | 9 |
| 2026-06-10 06:17 | completed | 1015 | 1015 | 0 | 46 |
| 2026-06-09 06:02 | completed | 1015 | 949 | 66 | 9 |
| 2026-06-08 06:36 | completed | 1015 | 943 | 72 | 9 |
| 2026-06-07 06:16 | completed | 1015 | 953 | 62 | 10 |
Why scans fail
Error breakdown from the latest completed run. Failed brands get one calm retry (lower concurrency, doubled timeout) before being counted here.
| Error | Brands |
|---|---|
| getaddrinfo EAI_AGAIN www.kroger.com | 1 |
| getaddrinfo EAI_AGAIN www.farmsteadapp.com | 1 |
| getaddrinfo ENOTFOUND www.drizly.com | 1 |
| getaddrinfo ENOTFOUND www.kfrancisbeauty.com | 1 |
| getaddrinfo EAI_AGAIN www.yourparade.com | 1 |
| getaddrinfo ENOTFOUND www.takecareof.com | 1 |
| getaddrinfo ENOTFOUND www.staub-usa.com | 1 |
| getaddrinfo ENOTFOUND www.parts.toyota.com | 1 |
8 brands failed their last 3+ consecutive scans (of 7 runs examined). Persistent failers are reviewed for removal or reclassification — a site that blocks our scanner is recorded as a finding, not silently dropped.
Care/of7/7 runsgetaddrinfo ENOTFOUND www.takecareof.com
Drizly7/7 runsgetaddrinfo ENOTFOUND www.drizly.com
Farmstead7/7 runsgetaddrinfo EAI_AGAIN www.farmsteadapp.com
Kosas7/7 runsgetaddrinfo ENOTFOUND www.kfrancisbeauty.com
Parade7/7 runsgetaddrinfo EAI_AGAIN www.yourparade.com
Staub7/7 runsgetaddrinfo ENOTFOUND www.staub-usa.com
Toyota Parts7/7 runsgetaddrinfo ENOTFOUND www.parts.toyota.com
Kroger3/7 runsgetaddrinfo EAI_AGAIN www.kroger.com
False-positive handling
- Two-scan confirmation: inference-based changes (HTTP verdicts, WAF/CDN, structured data) must appear in two consecutive daily scans before publishing. robots.txt diffs publish immediately because they are literal text changes. Details on /methodology.
- Scanner failures are never brand changes: timeouts, HTTP 429s, and network errors are recorded as inconclusive and excluded from the changelog.
- Policy vs enforcement separation: a WAF block is published as
restricted, never conflated with a robots.txtblocked. - Confidence labels: changelog entries are tagged high / medium / low confidence based on how the signal is measured.
Corrections log
No manual corrections issued yet. When we correct published data, the entry appears here permanently: date, brand, what was wrong, and what changed.
Dispute this data
If you represent a brand and believe a published data point is wrong:
- Email hello@arcreport.ai with subject
[DATA DISPUTE] your-domain.com(the link pre-fills the required fields). - We re-scan the brand out of band within 2 business days and compare against your evidence.
- If we were wrong, we fix the data, add a permanent entry to the corrections log above, and reply with what changed. If the data stands, we reply with the raw scan evidence so you can reproduce it.
Disputes never silently edit history: corrected values appear in the changelog like any other change.