Growth channels & retention — test with discipline, read cohorts
🎯 Goal: Learn to test growth channels with discipline (one at a time), read cohort retention to see if customers stay, and understand the referral loop that makes growth feed itself.
There are dozens of growth channels (ads, SEO, KOLs, referrals, content...) but you can’t run them all. The discipline: test ONE channel at a time, measure clearly, then scale what works. But buying customers is pointless if they don’t stay — so you must read cohort retention: track a group entering the same month and watch what % stays active over the months. Strongest of all is the referral loop: happy customers bring new ones, so growth feeds itself instead of buying all-new customers.
Lesson content
🎯
Retention is the foundation, not Acquisition. Many think growth = buying tons of customers. But if they arrive and all fall out, every ad dollar pours into a leaky bucket. Right order: first make customers stay (flat retention), then speed up buying. A cohort that "flattens" (doesn’t fall to zero) signals product–market fit: there’s always a retained core, the base for compounding over time.
🧭
The disciplined channel-testing process — 5 steps: 1) Pick 1 channel that fits your customers (don’t spread across 5 at once). 2) Set a hypothesis: who sees it, why they click/buy. 3) Run long enough for trustworthy numbers, measure cost per customer (CAC) and customer quality. 4) Decide: if good, scale (double down); if bad, kill it and try another. 5) Read cohorts in parallel: do this channel’s customers stay? A cheap channel whose customers churn fast is worse than a pricey one whose customers stay.
🧰
TOOL — Cohort retention table. Track a group entering at Month 0, watch % still active over months — a flattening line is good, falling to zero is leaky.
Cohort
Month 0
Month 1
Month 2
Month 3
Cohort A (healthy)
100%
62%
55%
52%
Cohort B (leaky)
100%
30%
12%
4%
A flattens toward ~52% → a loyal core, worth scaling. B falls to 4% → fix the product/experience BEFORE pouring in ad money.
🌏
Case study — Coolmate & the referral loop (step by step): 1) Coolmate is known for a 50%+ repeat rate (estimated) — that is healthy retention. 2) Healthy retention makes marketing far more efficient: it isn’t all new-customer buying every month. 3) Add a good referral loop (referral codes, customers inviting customers) → each new customer brings more → cheap, durable growth. 4) Conversely, if retention leaks but you still run heavy ads, customers arrive and leave at once → the more you run, the more you lose. Lesson: patch retention first, amplify later.
⚠️
4 growth traps: • Spreading money across 5 channels at once → noisy signal, can’t tell which truly works to double down. • Chasing only cheap CAC while ignoring quality: cheap customers who churn are useless. • Scaling while retention is still leaky: amplifies losses, pours money into an empty bucket. • Forcing referral: big rewards on a product not worth recommending only attract deal-hunters who churn fast.
✅
Channel-test & cohort checklist: ① Am I testing ONE channel at a time with a clear hypothesis? ② Am I measuring CAC AND customer quality (do they stay)? ③ Does my cohort line flatten or fall to zero? ④ If it falls, what product/experience will I fix BEFORE raising ad spend? ⑤ Is the product good enough that customers refer it voluntarily?
Practice exercise
🔬 APPLIED EXERCISE: Pick one growth channel to test first for your idea. Write the hypothesis (who sees it, why they’d click) and the metric you’ll measure (CAC, conversion rate). Then sketch a 4-month assumed cohort table and ask: does my retention flatten or fall to zero? If it falls, name one thing to fix BEFORE pouring in ad money.
Worked template: FILLED EXAMPLE — TikTok for a study app: hypothesis students watch study-tip videos → install. Measure views→installs (CAC). Assumed cohort: M0 100% → M1 40% → M2 18% → M3 8% → falls near zero = leaky retention. Fix first: add a 7-day learning path + reminders so customers have a reason to return before raising ad spend.
Quick quiz
1. What is the disciplined way to test growth channels?
→ Test one channel at a time, measure clearly, then scale what works
Channel discipline: test one channel at a time, measure clearly, then scale what works — don’t spread thin.
2. A "flattening" cohort retention curve (not falling to zero) usually signals what?
→ Product–market fit: a group of customers stays steadily
A flattening cohort curve (not falling to zero) signals product–market fit: a retained core stays.
3. Why is pouring in heavy ad spend while retention is leaky dangerous?
→ Because customers arrive and leave at once — the more you run, the more you lose
Heavy ads on leaky retention means customers arrive then leave — the more you run, the more you lose.
4. What does Coolmate’s "50%+ repeat rate" show?
→ Healthy retention, making marketing more efficient since it isn’t all new-customer buying
Coolmate’s 50%+ repeat rate is healthy retention, making marketing efficient since it isn’t all new-customer buying.
5. What advantage does a good "referral loop" give growth?
→ Each new customer brings more, making growth cheap and self-feeding
A good referral loop makes each new customer bring more — cheap growth that feeds itself.
6. When comparing two channels, why is "cheap CAC" not enough to call it good?
→ Because you must also check customer quality — cheap customers who churn are useless
Cheap CAC isn’t enough: weigh customer quality — cheap customers who churn are worse than pricey ones who stay.
Advanced
A deeper framework
At an advanced level, every growth decision reduces to one inequality: LTV > CAC (a customer’s lifetime value must exceed the cost to acquire them). Retention is what lifts LTV: the longer customers stay and the more they buy, the higher LTV — so fixing retention raises LTV without spending more on acquisition.
That’s why reading cohorts matters before scaling: a flattening curve (like Cohort A) means a retained core → real LTV → worth raising Acquisition. A curve falling to zero (Cohort B) means thin LTV → raising Acquisition only amplifies losses. The referral channel is a special "weapon" because it both lowers CAC and usually brings high-quality customers (friend referrals).
Cohorts decide whether to raise Acquisition spend
Cohort A: 100→62→55→52%
~52% core stays → real LTV, worth scaling
Cohort B: 100→30→12→4%
Nearly all churn → scaling = multiplying losses
Current CAC
~120,000₫ per customer
LTV of Cohort A vs B
A is many times B because customers stay longer
At the same CAC, only Cohort A yields durable LTV > CAC. Raising Acquisition into B pours money into an empty bucket.
Common trap: Splitting budget evenly across 5 channels at once: sales tick up but the signal is noisy — you can’t tell which to double down on or kill. Lack of channel discipline makes every measurement useless.
Advanced questions
1. You run 5 channels at once with a small budget split evenly; after a month sales tick up slightly but you can’t tell which channel did it. What’s the core mistake?
→ Lack of channel discipline: spreading thin so you can’t measure which channel truly works to scale
Discipline is testing one channel at a time, measuring CAC and quality, then scaling the winner. Splitting evenly across 5 muddies the signal — you can’t tell which to double down on or kill.
2. Cohort A: 100→62→55→52%. Cohort B: 100→30→12→4%. Which judgment is CORRECT when deciding whether to raise Acquisition spend?
→ A shows flattening retention (a PMF sign) so it’s worth scaling; B leaks, and scaling pours money into an empty bucket
A flattens toward ~52% — a retained core, a PMF sign, so more Acquisition can profit. B falls to 4% (leaky); scaling only amplifies losses — fix retention first.
3. Why is the referral channel often especially valuable compared to paid ads?
→ Because it both lowers CAC and usually brings high-quality customers (trusted friend referrals who stay longer)
Referral lowers CAC (existing customers bring new ones) and usually brings high-quality customers from friend trust — but it only runs when the product is good enough to be worth recommending.
🎯 Real-life mission
REAL-LIFE MISSION: Pick one growth channel to test first for your idea. Write the hypothesis (who sees it, why they click) and the metric you’ll measure (CAC, conversion). Then build a 4-month assumed cohort retention table (Month 0..3) and answer: does your line flatten (PMF) or fall to zero? If it falls, name one thing to fix BEFORE pouring in ad money.