Lesson 78.3 gave every send a job. This lesson checks whether the job got done — without vanity math and without clinging to readers who have already left emotionally. Email measurement is small and honest: a handful of signals read monthly, plus deliverability hygiene that keeps the inbox door open. Retention here means deserving the next open, not preventing the next unsubscribe.

Six signals that matter, one that misleads

• Delivered: emails accepted by the receiver, net of bounces. The floor of everything — if delivery slips, nothing downstream is readable.
• Clicks (and click rate): readers who followed the one action to the destination. The closest proxy to "this issue was useful enough to continue." Track unique clicks per send and clicks to the primary action separately from footer-link noise.
• Replies: readers who answered. The highest-quality signal in a small list — a reply means the message reached a human who cared enough to write back. ByeBuy's "reply with what you are building" exists to manufacture exactly this.
• Conversion / activation: the job completed — booking, order, template used, first brief finished. Ties the send to the distribution path from Class 67 (exposure → action → return), not to inbox theater.
• Unsubscribes: readers who used the honorable exit. Read as feedback per send: a small steady rate is healthy pruning; a spike names the issue that broke the promise.
• Engaged readers: the humans behind the averages — how many opened or clicked in the last 30–60 days, and who they are. A list of 800 with 300 engaged beats 5,000 with 200 silent.

Open rates are the misleading one. Apple Mail privacy preloading, image blocking, and forwarded rendering mean an "open" may be a bot prefetch and a missing "open" may be a reader who read the plain text. Use opens directionally at best; decide on clicks, replies, conversions, and unsubscribes.

Read the funnel per send: delivered low → deliverability or list problem; delivered fine but no clicks → promise or middle problem (wrong job, weak demonstration); clicks but no conversion → destination problem (Lesson 72.2 match failure); conversion without return → product or follow-up problem. Never celebrate the first number and ignore the rest — the Part XIV rule holds in the inbox too.

Deliverability, in plain language

Deliverability is whether mailbox providers place your mail in the inbox rather than spam or nowhere. Providers judge you as a sender reputation: do real people want this, and do you behave cleanly? The practices are unglamorous and durable:

• Permission only — never add buyers, attendees, or scraped addresses without a scoped yes (Lesson 78.1).
• One clear unsubscribe in every issue, honored within days; a visible exit prevents spam complaints, which damage reputation far more than unsubscribes do.
• Consistent sending identity — same from-name, same domain, authenticated (SPF/DKIM/DMARC via your provider's current setup guide), same weekly rhythm rather than bursts.
• Clean list — remove hard bounces promptly; re-engage or release the long-silent (Lesson 78.3) instead of shouting louder.
• Relevant content — the promise kept. The fastest way to the spam folder is a bored list that deletes unopened, which providers read as "unwanted."

Tool layers change — provider dashboards, authentication screens, spam-checker scores — so link your platform's current deliverability docs when you implement. The durable idea never changes: wanted mail from a consistent, permissioned sender reaches the inbox; clever mail from a resented sender does not.

The monthly review: keep, change, stop

Once a month, 45 minutes, same template. Pull the last four to five sends plus the welcome sequence's first-value rate (how many new subscribers completed message 2's action). Answer per recurring send, not per anecdote:

# EMAIL-REVIEW.md — [Month], list size: ___ / engaged (60d): ___

## Strongest useful issue
- Subject + job: ___ / clicks: ___ / replies/themes: ___ / conversions: ___

## Reply + theme scan
- Recurring questions (candidates for 77.4 log or next educational): ___
- Broken links/destinations fixed: ___

## Segments needing a different message
- New vs active mismatch seen: ___ / fix: ___

## Inactive readers
- Silent 60–90d count: ___ / re-engagement sent: Y/N / released: ___

## Deliverability check
- Bounce/spam-complaint trend: ___ / auth + identity unchanged: Y/N

## Decisions
- KEEP: ___ (why — evidence)
- CHANGE: ___ (what changes next send + review date)
- STOP: ___ (what we retire and what replaces the slot)

Worked mini-example — ceramics shop March: strongest issue was the educational shelf guide (41 clicks, 9 replies naming favorites, 6 reservations downstream). Launch had clicks but weak conversion — destination was the homepage, not the booking page; CHANGE: exact booking link next release. Re-engagement released 60 silent readers, list shrank 12%, engaged share rose; STOP: the monthly "shop news roundup" nobody clicked, replaced by a second care-tip educational. Deliverability steady, zero spam complaints. Smaller list, fuller Saturdays.

Finish line: an EMAIL-REVIEW.md with the month's evidence, reply themes, deliverability line, and explicit keep/change/stop for each recurring email.

Verify quickly: for every KEEP, point to clicks, replies, or conversions — not opens or "felt good." For every STOP, name what takes the slot so the calendar does not refill with drift.

Common failure mode: retention by guilt — hiding the unsubscribe, sending "we miss you!!!" weekly to the silent, celebrating list size while engaged readers decay. Respectful retention grows the readers who act and releases the rest. Attention kept against its will converts to spam complaints.

Check your understanding

1. Which six signals matter, and why are open rates only directional? 2. What five practices earn inbox placement, and why do spam complaints hurt more than unsubscribes? 3. What must a monthly review decide for every recurring send?

Next

Class 78 closes here: permission earned, welcome routed, four jobs calendared, measurement honest. You now own one direct relationship route that no feed can repossess. Class 79 widens the portfolio again — partnerships, marketplaces, and product-led loops — where borrowed trust and built-in intent meet the audience, search, social, and email paths you already run.