Deloosh does not show an open rate as an engagement or performance metric because opens stopped being reliable evidence. It never uses opens for recommendations, decisions, triggers, segments, or sending-health assessments. When the assistant surface is available, you can still ask for the recorded open count on an email; Deloosh will disclose it with a mandatory inaccuracy warning.
Opens are dead evidence
Apple Mail Privacy Protection and Gmail image prefetch can load tracking images on the recipient’s behalf, whether or not a person reads the message. The result is an open count inflated by software, and inflated unevenly across your list depending on which mail app each person uses. You cannot treat it as proof of attention, compare it cleanly across audiences, or act on it.
Deloosh therefore keeps opens out of every evidence and decision surface. Opens do not support a recommendation, qualify engagement, trigger a play, select an audience, or affect the sending-health light. Recorded clicks and downstream behavior are the engagement evidence instead.
Asking for your open count
Open counts are not hidden. When the assistant surface is available, ask for the count on an email and it will return the recorded number. It always includes this caveat: the count is likely inaccurate because Apple Mail Privacy Protection and Gmail image prefetch can auto-load images without a human opening the message.
That is disclosure, not decisioning. The assistant will not turn the count into engagement evidence or recommend an action from it, even if you ask.
The disclosure exists only through the assistant surface. That surface is currently behind the default-off HARNESS_ENABLED launch gate, so there is no live open-count view while the gate remains off. Enabling it makes the caveated count available on request; it does not add an open-rate dashboard or weaken the evidence ban.
What Deloosh reasons from instead
- Recorded clicks — activity observed on a link, rather than an open generated by privacy prefetch. Raw clicks remain unclassified until the future human-versus-scanner classifier ships.
- Bounces — the mailbox rejected the message. Hard evidence about the address.
- Complaints — a complaint event recorded by the mailbox provider. Evidence that the recipient must be suppressed.
These are recorded activity, not privacy-prefetch noise. They do not all drive the same features: clicks do not feed the sending-health light, suppression list, or pacing projection. The health light has its own seven signal families.
One honest caveat: a recorded click is a click Deloosh observed. It is not yet classified as human-versus-scanner, so the product treats clicks as recorded activity and does not overclaim them.
You are never charged for opens, either
Opens are not a billing meter and are not gated by any plan. Nothing in Deloosh’s pricing depends on them, so there is no commercial reason for the product to inflate or feature them.
What you lose, honestly
If you are used to reporting an open rate, you will not find an open-rate dashboard or a decision metric here. Once the assistant surface is available, you may ask for the recorded count, but it arrives with the inaccuracy caveat and is never treated as performance evidence. The product uses recorded events while preserving their evidence boundary: a click is observed activity, not yet proof that a human produced it.
This is the same principle behind refusing to invent claims in drafts: see TRUST-1, What Deloosh does on its own — and what it always asks first.