Retention conversations end in numbers — but most subscription numbers quoted online are costumes. This lesson defines each metric plainly, one at a time, works one transparent hypothetical by hand, and leaves you with a scorecard that points at the next product decision instead of the next pitch deck.

The vocabulary, one at a time

• Monthly recurring revenue (MRR): what active subscriptions are worth per month if everyone stays. A snapshot, not cash in hand.
• Annual recurring revenue (ARR): MRR × 12 — a run-rate shorthand. It annualizes today's snapshot; it does not mean the money is collected or committed.
• Gross margin: what remains after direct delivery costs — payment-provider fees, model and hosting spend, direct support time — divided by revenue. High margin means growth helps; low margin means growth hauls cost with it.
• Acquisition cost (CAC): what you spend to win one paying customer — ads, content time, sales effort, trial support. Blended averages lie; know it per channel.
• Lifetime value (LTV): expected gross profit from one customer before they churn. Roughly: monthly price × margin × expected months retained. An estimate, not a payment.
• Payback period: how many retained months it takes for one customer's margin to repay their CAC. Short payback forgives mistakes; long payback punishes churn.
• Cohort: customers grouped by when they started — January signups, trial-batch 12. Cohorts reveal whether changes helped, where blended averages only soothe.

For how these map to real billing constructs — subscriptions, meters, trials — keep the Stripe Billing documentation nearby, and compare packaging choices against the Stripe recurring pricing models. Mechanics confirm; they do not decide.

One transparent hypothetical (illustrative assumptions, not a forecast or provider quote)

Thirty active customers × $30/month = $900 MRR — before anything is paid to anyone. That "before" is the entire lesson. Work it by hand:

• Provider fees (illustrative ~5% assumption — verify current provider pricing): $45. Model and hosting (assumed ~$6 per active customer): $180. Direct support (assumed ~$3 per customer): $90. Direct costs: $315. Gross profit: $585. Margin: 65%.
• Suppose winning each customer cost $60 in content and onboarding time. CAC: $60. Payback: $60 ÷ ($30 × 65%) ≈ 3.1 retained months. A customer who churns in month two never repays acquisition.
• Suppose average retention is 10 months. LTV ≈ $30 × 65% × 10 = $195. LTV-to-CAC ≈ 3.3:1 — workable, not luxurious.

Now feel the sensitivities. Raise the price to $36 with no churn change: MRR becomes $1,080, margin widens, payback shortens to ~2.6 months. Let model costs double to $12 per customer instead: direct costs jump to $495, margin collapses to 45%, payback stretches past 4 months, and every new customer hauls more weight. Lose two extra customers a month to churn: MRR stalls even as acquisition spending continues, and LTV falls faster than any price tweak repairs — because churn shortens every lifetime at once. One small table, three levers, no theater: price moves the top, cost moves the middle, retention moves everything.

What ARR hides

ARR is MRR wearing an annual coat: $900 MRR is "$10,800 ARR." The number sounds collected. It is not. It is today's snapshot multiplied by twelve, assuming nobody leaves, nobody downgrades, and costs stay flat.

A high top line can hide three uncomfortable truths. High delivery cost: $10,800 of run-rate at 30% margin is a smaller business than $6,000 at 80%. High churn: a growing customer count with a leaking bucket means acquisition spending sprints to stand still — cohorts tell the truth the total hides. Collected-versus-committed: monthly plans can vanish next month; only cash, annual prepay, and retained cohorts are real. Quote ARR as shorthand among people who know the definition, and always beside margin, churn, and cohort retention. Alone on a slide, it is a costume.

Exercise: build the subscription scorecard

Create SUBSCRIPTION-SCORECARD.md. 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 as 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: ___% / CAC this month: ___ / payback (months): ___ / rough LTV: ___

## 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 ___."

Worked mini-example — Desk month one: 30 start, 6 new, 24 retained, 6 cancelled (4 voluntary, 1 failed-payment recovered, 1 inactive flagged for outreach). MRR $900 → $900 with price $30 flat; lost $180, gained $180, net zero — growth theater exposed. Direct costs $315, margin 65%, CAC $60, payback ~3 months. Cohort note: trial-batch signups retain 40% to week four versus 70% for demo-led teams — the door, not the price, is the bottleneck. Learning question: "Week-four activation wants the onboarding decision, so next I will inspect the 48-hour checklist from Lesson 81.4's log."

Finish line: a scorecard you can explain aloud — what each number means and which product decision to inspect next.

Verify quickly: read the scorecard to a peer and ask them to point at the riskiest assumption. If they point at ARR instead of margin, churn, or the cohort note, the page is still costumed — move the cohort and cost rows above the run-rate.

Common failure mode: theater math — quoting ARR as collected revenue, LTV as cash, or blended CAC across wildly different channels. Each flatters the present and starves the decision. Label estimates as estimates, split what matters, and let the learning question be the headline.

Check your understanding

1. Define MRR, gross margin, payback, and cohort — and why ARR alone misleads. 2. In the 30 × $30 hypothetical, what happens to margin and payback if per-customer costs double? 3. What must SUBSCRIPTION-SCORECARD.md contain, and why does it end in a question?

Next

Class 81 closes with economics you can defend. Class 82 packages a different asset — your own expertise — into services, workshops, and courses that teach before they scale.