Lesson 81.5 defined every subscription metric plainly and warned that top-line numbers costume weak economics. This appendix is the monthly instrument: customers, recurring revenue, direct costs, cohort note, and one learning question — plus a filled 30 × $30 example you can copy.

The blank scorecard

One page, updated monthly, ending in a question — never in a boast.

# SUBSCRIPTION-SCORECARD.md — [Project] — Month: ___

## Customers
- Start of month: ___ / new (paid): ___ / retained: ___
- Cancelled: ___ / end of month: ___
- Voluntary cancels: ___ / failed-payment cancels: ___ / inactive-but-paying: ___

## Revenue (recurring only; one-time sales listed separately)
- MRR (end of month): ___ / ARR run-rate (MRR × 12, labelled run-rate): ___
- New MRR from new customers: ___ / MRR lost to cancels/downgrades: ___
- Net MRR change: ___

## Direct costs (monthly)
- Provider/billing fees: ___ / model + hosting: ___
- Direct support time valued at: ___ / total: ___
- Gross profit: ___ / margin: ___%

## Cohort note (the truth serum)
- Cohort watched: ___ / retained after 1 / 3 / 6 months: ___ / ___ / ___
- Lesson: ___

## One learning question (required)
- "The number that most wants a product decision is ___ because ___,
   so next I will inspect ___."

Each section answers one question. Customers separates growth (new) from health (retained versus cancelled) and splits cancels by cause — voluntary churn is a product signal, failed-payment churn is a billing signal, inactive-but-paying is a trust debt you should resolve before it becomes a chargeback. Revenue keeps one-time sales out of MRR so launches cannot disguise a leaking bucket. Direct costs hold the three honest categories: what the payment provider takes, what delivery costs (model, hosting, tools), and what support time is worth. Cohort defeats blended averages. Learning question converts arithmetic into an assignment.

Filled example: 30 customers × $30 (illustrative hypothetical, not a forecast)

Research Desk, October, founder-filled in twenty minutes:

# SUBSCRIPTION-SCORECARD.md — Research Desk — Month: 2026-10

## Customers
- Start: 30 / new (paid): 6 / retained: 24
- Cancelled: 6 / end: 30
- Voluntary: 4 / failed-payment: 1 (recovered) / inactive-but-paying: 1 (flagged)

## Revenue
- MRR (end): $900 / ARR run-rate: $10,800 (run-rate, not collected)
- New MRR: +$180 / lost MRR: −$180 / net change: $0

## Direct costs
- Provider fees (~5%): $45 / model + hosting: $180 / support (~$3/customer): $90
- Total: $315 / gross profit: $585 / margin: 65%

## Cohort note
- Cohort watched: September trial batch vs. demo-led teams
- Retained to week four: 40% vs. 70%
- Lesson: the door, not the price, is the bottleneck

## One learning question
- "The number that most wants a product decision is week-four activation
   (40% trial vs. 70% demo) because acquisition holds steady while retention
   leaks, so next I will inspect the 48-hour onboarding checklist."

Read the story the numbers tell: customer count flat at 30, MRR flat at $900 — new revenue exactly replaced churned revenue. That is growth theater exposed by one subtraction. Margin at 65% is workable, but the cohort split says the next dollar belongs in onboarding, not acquisition. Nothing here needs a forecast; it needs a checklist fix.

Rules that keep it honest

• Fill it the same day each month before any new work. Missing months destroy the slope that cohorts need.
• Label estimates as estimates. Rough lifetime value and acquisition cost belong in the notes with their assumptions, not as headline facts.
• Quote run-rate as run-rate. "$10,800 ARR" without the words "run-rate" will be misread as collected cash within a quarter — by you first.
• When packaging or fee assumptions change, recheck them against current sources: the Paddle pricing strategy guide for value-metric and tier logic, and Gumroad pricing if a simple storefront handles your membership billing — provider fees move, and your cost row must move with them.

Verify fast

Hand the filled page to a peer and ask: what is the riskiest number, and what should I inspect next? If they point at ARR instead of margin, churn, or the cohort note, the page is still costumed — move the cost and cohort rows above the run-rate and re-test.