Lessons 86.1–86.4 built the machine: journeys mapped, segments defined, a two-week brief written, AI preparing under a human gate. Now the clock runs out on the campaign — and someone asks "so, did it work?" Most reports answer with volume: sends, opens, clicks. This lesson teaches the report that answers with a decision.

The vocabulary, in plain language

• Baseline: what happened before the campaign — the rate you compare against. No baseline, no verdict.
• Qualified lead: a lead that passed the fit check (Lesson 86.2) — right problem, reachable, real. Raw lead counts flatter; qualified counts inform.
• Conversion: the one primary event from the brief — bookings, purchases, signups. Counted once, defined in advance.
• Acquisition cost: total campaign cost divided by conversions — money, tool time, and human hours included. A campaign that "worked" at an unpayable cost did not work.
• Assisted conversion: a conversion whose journey touched multiple messages — the checklist, both emails, the founder's reply. Record every touch; resist crowning one hero.
• Incrementality: the conversions that would *not* have happened without the campaign — the lift above baseline, not the total. Ten bookings against a baseline of six means four incremental, not ten.
• Decision log: the written record of what you decided, why, and what you will test next. Memory lies; logs compound.

The mental model: a report is a doctor's rounds note, not a trophy shelf. It says what changed, what you think caused it, how sure you are, and what treatment comes next — in language the next reviewer can act on.

The weekly shape: five lines that decide

Every weekly report follows the same five moves:

1. Result. One sentence with the primary metric against baseline: "6.1 bookings per 100 sends vs. baseline 2.0." Numbers first, adjectives never. 2. Plausible cause. Your best explanation, stated as hypothesis: "the fitted-checklist promise plus the Day-6 worked example likely did it — bookings spiked the mornings after touches 1 and 3." 3. Evidence. What supports the cause — and what complicates it: touch-day spikes, reply quotes naming the example, plus the caveat that a partner mention mid-week may have assisted. 4. Uncertainty. What you do not know, stated plainly: "small sample (312 sends); partner overlap unmeasured; no holdout." Uncertainty is not weakness — it prices the next test correctly. 5. One next test. The single experiment the report buys: "hold out 20% next fortnight to isolate the Day-6 example's effect; kill threshold under 3.0 per 100."

Add acquisition cost every week, even when it hurts: "€41 per booking all-in (tool €90 + 6 owner-hours)." A metric without a cost is a hobby. And keep the decision log rolling — date, decision, reason, test — so month three inherits month one's learning instead of re-arguing it.

Exercise: report a hypothetical campaign

Create MARKETING-WEEKLY.md for the Lesson 86.3 brief's second week. Invent honest numbers — including an imperfection — and practice the five-move shape before real money is involved.

# MARKETING-WEEKLY.md — [Campaign] — Week ending ___

## Result (primary metric vs baseline)
- ___ per 100 vs baseline ___ (sends: ___ / conversions: ___ / qualified: ___)

## Plausible cause
- Best explanation (hypothesis, one paragraph): ___

## Evidence
- For: ___ (touch-day timing, replies, assisted touches recorded)
- Against / complicating: ___ (overlap, seasonality, small sample)

## Uncertainty
- What we cannot claim: ___ / sample size ___ / missing measurement ___

## Cost
- All-in spend: ___ / cost per conversion: ___ / owner hours: ___

## Decision log
- Date ___: decided ___ because ___ / next test ___ with kill threshold ___

Worked mini-example — ByeBuy's fitted-checklist fortnight: result "6.1 bookings per 100 (19 bookings / 312 sends) vs. 2.0 baseline; 14 of 19 qualified on fit check." Cause: "the Day-6 anonymized fitted example did the heavy lifting — 9 of 19 booked within 24 hours of it; three replies named it." Evidence for: touch-day spikes plus quotes; complicating: a partner linked the worksheet mid-week (~40 extra visits, unattributed). Uncertainty: small sample, no holdout, partner assist unmeasured — "claim lift, not proof." Cost: "€780 all-in, €41 per booking, 6 owner-hours." Decision: "repeat a fortnight with a 20% holdout; kill under 3.0; owner drafts the holdout plan by Friday."

Finish line: a MARKETING-WEEKLY.md with all five moves, an honest uncertainty line, all-in cost, and one logged next test with a kill threshold.

Verify quickly: hand the report to someone outside marketing. If they can state the decision and the test it bought — without your help — the report decides. If they admire the opens chart and ask "so what now?" — rewrite.

Common failure mode: the vanity weekly — opens up, clicks up, screenshots of spikes, no baseline, no cost, no test. Everyone feels good; nothing changes; the same campaign runs unexamined for a year. Baselines and kill thresholds are the cure.

Check your understanding

1. Why do qualified leads and incrementality matter more than raw sends, opens, or total conversions? 2. What are the five moves of the weekly report, and what does each contribute? 3. What belongs in a decision log, and why does month three need month one's entries?

Next

Class 86 ends here: you can trace a journey, segment it, brief a bounded test, gate AI preparation, and report a verdict. Class 87 turns to the people already asking for help — support as a product sensor, a triage layer, a CRM with memory, lifecycle follow-up without spam, and the loop that turns evidence into fewer tickets.