Lesson 86.2 sorted your audience into segments that each deserve a different next step. A campaign is what happens when you serve one segment, one promise, for a bounded time — and learn something. Without that bound, "the campaign" becomes permanent background mailing that nobody reviews and nobody can kill.

The vocabulary, in plain language

• Objective: the one outcome this campaign exists to move — e.g. "book 15 walkthroughs from worksheet finishers." If everything is the objective, nothing is.
• Audience: the one segment, named exactly as in Lesson 86.2 — signal in, stop out.
• Offer: the concrete value — the teardown, the office-hours seat, the trial extension. The offer is what the reader gets, not what you want.
• Message: the one promise in the reader's words — what changes for them, and what it costs in time or money.
• Channel: where the message travels — email, a lesson page box, a partner slot, a call. Each channel has its own manners.
• Asset: the actual object — the email draft, the landing paragraph, the one-page PDF. A brief without assets is a wish.
• Cadence: how many touches, on which days. Two weeks, three touches, named dates — not "ongoing nurture."
• Control: the baseline you compare against — last period's rate, a holdout half of the segment, or simply "no campaign." Without a control, every result is a story.
• Postmortem: the 30-minute review when the clock runs out: what happened, what you believe caused it, what you will test next.

The mental model: a campaign is a laboratory flask, not a loudspeaker. Small, labeled, sealed for two weeks, then opened and inspected. The loudspeaker model — permanent volume, no label, never inspected — is how budgets evaporate.

The one-each rule

Every two-week campaign gets exactly one objective, one audience, one promise, one primary metric, and one learning question. Write all five on the brief before a single asset is drafted:

• One objective: "book walkthroughs" — not "book walkthroughs and grow the list and test pricing."
• One audience: "worksheet finishers, stalled 7+ days" — not "everyone who ever downloaded anything."
• One promise: "in 25 minutes, leave with a deployment checklist fitted to your stack."
• One primary metric: "walkthroughs booked per 100 sends." Secondary numbers are context, not verdicts.
• One learning question: "does a fitted-checklist promise beat a generic demo invite for stalled finishers?"

Why so strict? Because a campaign testing three promises on four audiences across two channels teaches nothing when the numbers move — you cannot tell which change mattered. The one-each rule buys you causal clarity at the cost of temporary narrowness. After the postmortem, you may widen.

Channels and cadence follow the segment's manners. Email suits the stalled finisher (they gave an address for exactly this). A lesson-page box suits the anonymous reader (no permission yet — offer, don't chase). A partner slot suits borrowed trust (their audience, their rules, your fitted promise). State the channel per touch, with dates: "Day 1 email, Day 6 email, Day 11 final call." When Day 14 arrives, the campaign ends — even if it is working. Winners earn a *new* brief; they do not become permanent drips by inertia.

Exercise: write the two-week brief

Create CAMPAIGN-BRIEF.md. Fill it before drafting any asset — the brief constrains the work, not the reverse.

# CAMPAIGN-BRIEF.md — [Name] — [Dates, 14 days]

## The one-each block
- Objective: ___ (one outcome, counted how: ___)
- Audience: ___ (segment + signal in / stop out)
- Promise (reader's words): ___
- Primary metric: ___ (per 100 sends/views)
- Learning question: ___

## Offer + message
- Offer: ___ / Cost to reader (time/money): ___
- Message (≤ 40 words): ___

## Channel + cadence
- Touch 1 (day ___, channel ___): asset ___ / owner ___
- Touch 2 (day ___, channel ___): asset ___ / owner ___
- Touch 3 (day ___, channel ___): asset ___ / owner ___
- Suppression honored: ___ (unsubs, converted mid-flight, bounced)

## Control + review
- Control/baseline: ___ (last rate ___ / holdout ___)
- Postmortem date + owner: ___ / decision it will inform: ___

Worked mini-example — ByeBuy: objective "book 15 walkthroughs"; audience "deploy-checklist finishers stalled 7+ days"; promise "leave with a checklist fitted to your stack in 25 minutes"; metric "bookings per 100 sends"; question "fitted-checklist vs. generic demo invite." Touches: Day 1 email with fitted promise, Day 6 email with one anonymized fitted example, Day 11 plain-text final call from the instructor. Control: last quarter's generic-invite rate of 2 per 100. Postmortem on Day 15, owner named.

Finish line: a CAMPAIGN-BRIEF.md with the one-each block, three dated touches, suppression, control, and a scheduled postmortem.

Verify quickly: cover the assets and read only the one-each block to a peer. If they can predict what the emails say and how success is counted, the brief works. If they ask "who is this for, actually?" — narrow the audience.

Common failure mode: the forever campaign — no end date, three objectives, daily sends to the whole list. Numbers drift, nobody reviews, unsubscribes climb, and the team calls it "nurture." Kill it; write a two-week flask instead.

Check your understanding

1. What are the five "one-each" items, and why does each campaign get only one of each? 2. What is a control, and what goes wrong when a campaign has none? 3. Why must a winning campaign still end on Day 14 instead of running forever?

Next

A brief constrains the promise — but who writes the emails, and who approves the claim inside them? Lesson 86.4 draws the line: AI prepares drafts, research, and summaries at speed; a human owns claims, budgets, partners, audiences, and every brand-sensitive reply.