Feature 01
Target43%
Verified human open rate, the floor we intend to underwrite
Apple opens your email before the customer does. Your dashboard counts it as a read.
Apple Mail Privacy Protection fetches every image the moment a message is delivered, for every recipient, whether or not anyone looks at it. Security gateways sweep the same pixels seconds later. In a raw log you cannot tell either one from a person. That is why a large share of every open rate in this industry is fiction, and why nobody underwrites one.
This is a target, not a measurement
A contractual floor per campaign class, to be underwritten with the Deliverability SLA. It is not a measured average across customers, and it will vary by campaign class and list quality.
The difference
Everywhere else
Counts every pixel fetch as an open. Sends from mailboxes rented on a shared API, so there is no way to tell a person from Apple's relay or a security sweep. The number goes up and nobody can say what it means.
Wysera
Runs its own mail cluster, so the signals that give a fetch away are in the sending path. Sorts every fetch into five verdicts and reports only the ones a person caused. Then puts a contractual floor on that number, credited back when it is missed.
The sort
What gets thrown away is what makes it count.
All five of these arrive looking the same. Telling them apart is the product. It is also the only reason anyone can put a floor under the number and mean it.
Image and link requests against a sent message, at delivery and for as long afterwards as they keep arriving.
- Human openCounted
A person opened the message. This is the number the rate is built from.
- ReplyCounted
Someone wrote back. A reply always counts, whatever the pixel did.
- Proxy openCounted apart
Probably a person, behind something that rewrites the request. Counted, but kept apart from the human figure.
- Privacy-relay prefetchDiscarded
Apple Mail Privacy Protection fetching images at delivery for every recipient, whether or not anyone looked.
- Gateway scanDiscarded
A security gateway sweeping pixels and links seconds after delivery.
What is left once the fetches nobody caused are removed. In a raw log all five of these look identical.
The mechanism
How it actually works.
Six mechanisms, each one a thing that runs. A capability claim without one is a slogan.
- 01
Own the sending path
A mail cluster with per-domain caps, send pacing and a bounce circuit breaker. Not a resold API.
- 02
Warm and grade every mailbox
A daily ramp sets the send limit, and each mailbox is regraded from its own send history.
- 03
Classify every fetch
Classifier v3 scores against header self-consistency, network ranges and gateway MX patterns.
- 04
Count only the real ones
Human and proxy opens are counted separately. Privacy-relay prefetches and gateway scans are discarded. A reply always counts.
- 05
Re-score history when rules improve
Verification is computed on read rather than frozen at delivery, so the whole archive can be re-graded when the rules get better.
- 06
Underwrite it
A verified-open floor per campaign class, credited back when missed. That is only possible because the number means something.
Questions
The questions that come first.
Why is 43% a target rather than a measured average?
Because it is a commitment, not a result. It is a contractual floor per campaign class, to be underwritten alongside the Deliverability SLA. It is not an average observed across customers, and it will vary by campaign class and list quality. Anywhere the figure appears on this site it carries that qualification in the same breath.
What is Apple Mail Privacy Protection doing to my open rate?
It prefetches images at delivery for every recipient, whether or not anyone opens the message. Security gateways do something similar, sweeping pixels and links seconds after delivery. In a raw log neither is distinguishable from a person opening the email, which is why a large share of every open rate in this industry is fiction.
What actually counts as a verified open?
A fetch the classifier attributes to a person. Human and proxy opens are counted separately, privacy-relay prefetches and gateway scans are discarded, and a reply always counts regardless of what the pixel did. The classifier scores against header self-consistency, network ranges and gateway MX patterns.
What happens to past campaigns when the classifier improves?
They are re-scored. Verification is computed when the number is read rather than frozen at delivery, so the whole archive can be re-graded when the rules get better. An open rate from last quarter is not stuck with last quarter's understanding of what a robot looks like.
Why does owning the mail cluster matter for this?
Because the signals the classifier needs are in the sending path. Wysera runs its own mail cluster with per-domain caps, send pacing and a bounce circuit breaker, rather than reselling a shared sending API. A rented seat on somebody else's infrastructure cannot tell you which fetches it caused.
Run it against your own sending.
Bring a campaign you have already sent. See how many of its opens survive the sort. The argument on this page is that the number is checkable, so check it.