7 of 11 pitch emails sent in September were opened and 0 answered; across 121 sends since July every reply came within 2.1 hours of its send, and 28 of 31 openers never replied

What an open is here

The pitch lever sends short, identity-transparent emails from our own mailboxes to named people at outlets the AI engines cite, one pitch per outlet and at most one follow-up. Open tracking is on for this campaign, so an open is the platform's pixel firing at least once for that message. A pixel fire can be a person reading or a mail client fetching images ahead of time, so an open here means the message reached a mailbox and something rendered it, and nothing more. A reply is a received message on the campaign from the pitched address. The by-touch read on 2026-09-06 set the shape of this ledger; today's read adds the three September cohorts and the question of what an open that never answers is worth.

The three September cohorts

7 of 11 opened is 63.6%, above the lifetime open share of 40.3% of people, and the reply count is 0. The two first-touch cohorts (8 emails) are each due one follow-up, on 2026-09-19 and 2026-09-20; the 3 second-touch rows have had their one follow-up and get nothing more. Every open in the table is older than the longest reply delay this campaign has ever produced, which is the next section.

When a reply arrives, on the whole record

3 replies in 121 sends is 2.5%. All 3 came from people who opened the email they answered, and the slowest arrived 2.1 hours after the send. The 11 September rows were read at 24 hours at the youngest, 11 times that delay. On this record a reply is a same-session event: the person opens, decides, and writes within about two hours, or does not write.

Opens against replies, lifetime

By opens on the email, and by touch

A second open on the same email does not move the reply share: 1 reply in 10 two-open records and 1 in 7 with three or more, against 1 in 25 single-open records. 14 people opened one email two or more times and never wrote. The second touch produced 2 of the 3 replies from 44 sends, the first touch 1 from 77, which is why the one follow-up stays in the sequence; it does not change what an unanswered open predicts.

People who received both touches

11 of 39 people opened both emails, 28.2%, and 2 of the 39 replied. An open on the first touch that did not answer was followed by an open on the second touch more often than by silence (11 against 6), so the open pixel is consistent per person; it is the reply that does not follow.

What this read does not show

An open count is a pixel count and cannot separate a person reading from a client prefetching; the same limit applies to every row equally, so the comparison between openers and repliers stands, but the 40.3% open share is an upper bound on people who read. 3 replies is too few to estimate a rate with any width; 2.5% is the record, not a forecast. The 6 records without a measurable open count are July follow-ups sent as a sequence step on the same lead record and one deleted record, as the 2026-09-06 read stated.

The next read, pre-registered

1. 1(A) On 2026-09-23: Of the 11 September rows, replies arriving more than 24 hours after the send they answer. Expected 0 (every reply on the record so far arrived within 2.1 hours of its send); falsified if >=1.
2. 2(B) On 2026-09-23: Replies from the 11 September rows to the touch already sent (a second touch sent on or after 2026-09-19 counts as its own send). Expected 0 to 1 (3 of 121 sends lifetime = 2.5 percent; 11 x 2.5 percent = 0.3); falsified if >=2.
3. 3(C) On 2026-09-23: Openers among the 11 who never reply, read at 2026-09-23. Expected at least 6 of 7 (28 of 31 lifetime openers never replied); falsified if <=4 of 7.

Any leg outside its statement is published as a falsification with the same script and a dated artifact.

Artifact

`content/news/open-no-reply-cut-2026-09-16.json` holds the three September cohorts (per row: touch, send stamp, hours since send at the read, opens, replied), the lifetime counts by touch, by opens on the email and for two-touch people, the reply delays, the three pre-registered statements with their thresholds, and 16 self-checks (0 failed), one of which asserts that no address, name or URL is in the file. It is produced by `content/news/open_no_reply_cut.mjs` from the same day's `pitch-touch-cut-2026-09-16.json` (`pitch_touch_cut.mjs`, 14 of 14 self-checks) and one read of the sending platform's sent and received lists.