Transparency report
How we turn raw signals into verdicts, what data touches our systems, and what email tracking cannot tell you. Published because trust requires receipts.
How we classify opens and clicks
Every engagement event (pixel fetch, link click) is analyzed by the Evidence Engine: it gathers evidence — client and network behavior, timing relative to send, presence of clicks — and assigns one of five verdicts with a confidence score (0.0–1.0) and a human-readable explanation log. Rules run in fixed priority order: AI_CRAWLER, then SECURITY_SCAN, then PRIVACY_PROXY, then HUMAN; if nothing matches reliably, the verdict is UNKNOWN.
The five verdict buckets
HUMAN
Likely a person in a real mail client: residential network, plausible timing, ordinary browser signals.
PRIVACY_PROXY
A privacy relay fetched the content for the recipient — e.g. Apple MPP. Excluded from human KPIs.
SECURITY_SCAN
A corporate gateway, antivirus or sandbox opened the message before any human could.
AI_CRAWLER
An AI assistant or search crawler fetched the content.
UNKNOWN
Not enough evidence. We report unknown instead of guessing.
What data we process
Tracking must answer to privacy law, so we minimize by default:
- Per-recipient tokens are 128-bit random identifiers — no names or email addresses inside tracking URLs.
- IP addresses are pseudonymized at ingestion; we do not store full IPs with events.
- Raw events are time-boxed by retention limits, then aggregated or erased.
- Consent records capture source and timestamp; recipients get a privacy portal for opt-out, export and erasure (GDPR Art. 15/17).
- Data is hosted in the EU; TraXmark is operated by TriStiX S.L. (Spain) under GDPR and ePrivacy.
TriStiX S.L. · Alicante, España
Honest limits
What email tracking cannot do — stated by us, not discovered by you:
- A pixel fetch is an open signal, never proof that a person read the email.
- Classification is probabilistic: confidence scores, not certainties. Some events stay UNKNOWN.
- Recipients who block images produce no pixel event at all — opens can be undercounted.
- Apple MPP pre-fetches images for everyone; open rates that ignore this are inflated by design.
Your rights
Data protection questions: privacy@traxmark.com. Security reports: security@traxmark.com. Full documents: privacy policy, DPA and terms of service.
FAQ
- Does TraXmark prove an email was read?
- No. TraXmark classifies engagement signals (opens, clicks) with evidence and confidence. A pixel fetch is a signal, never proof of reading — this limit is published in our transparency report.
- What happens when classification is uncertain?
- The event gets the UNKNOWN verdict. TraXmark does not guess: unknown events are reported as unknown and excluded from human KPIs.