Signal quality and limitations
Interpret likely-human, privacy-protected, scanner, self, unknown, location, and multi-recipient evidence correctly.
Elite stores the incoming request as a signal, then derives a classification with a confidence level and reason. This separation allows classification to improve without pretending the original network request said more than it did.
Classifications
| Classification | Meaning | Counts as likely engagement? |
|---|---|---|
| Likely human | Timing, client, and network evidence fit a reader and no dominant automated signature is present | Yes |
| Privacy protected | Known privacy proxy or preload requested remote content | No; shown separately |
| Security scan | Link or image was fetched by known automated security infrastructure | No |
| Sender/self | Activity appears to come from the sender's own preview, sent folder, or test | No |
| Unknown | Evidence is insufficient or conflicting | No |
| Implied by click | A likely-human click proves message access despite no direct human-classified pixel | Yes, with the reason preserved |
Privacy services and caches
Apple Mail Privacy Protection can download remote content privately and hide the recipient's IP. Gmail serves external images through Google's proxy. Other clients and gateways cache images. These systems may fetch early, hide the device or location, or suppress later requests.
Accordingly, the dashboard says likely opened, not “read,” and does not turn a proxy data-center city into the recipient's location.
Device, network, and location
Paid evidence views can show parsed client, browser, operating system, device class, network/ASN, and location when the request supports it. Treat these as attributes of the observed request. They may be unavailable, reduced by browser privacy controls, or belong to a proxy/VPN/security service.
Multiple recipients
One standard email body to several recipients normally contains one identical tracker. An event can be attributed to that group, not a named member. Cc and Bcc do not change the technical limitation. Use individualized campaigns for recipient-scoped tokens.
Why counts differ across pages
- Total signals includes accepted pixel requests across classifications.
- Likely opens includes human-classified or clearly implied engagement.
- Unique opens cluster repeat evidence; a timeline may show more raw events.
- Click rate uses likely-human clicks, excluding known scanner requests.
- Filters and date ranges can differ between pages.
- Provider sync and enrichment may finish after the initial request.
Decisions tracking should not make for you
Do not use open data alone for employment decisions, legal notice, medical urgency, consent, safety checks, or any situation requiring proof that a person read content. Use an explicit reply, authenticated portal acknowledgement, or appropriate signature workflow.