Growth Loops
Date: 2026-09-27
A funnel ends at the customer; a loop starts the next one from them. Whether a loop compounds or fizzles comes down to one multiplier — how many new customers each cohort produces — and that multiplier is almost always below one, which makes the loop an amplifier on paid acquisition rather than a replacement for it.
A growth loop is an acquisition system where the output of one cycle — a customer, a piece of content, a shared item — becomes the input that brings in the next.
The framing was popularised by a 2018 Reforge essay, Growth Loops are the New Funnels, by Brian Balfour, Casey Winters, Kevin Kwok and Andrew Chen.
Funnel versus loop
FUNNEL LOOP
spend → visitors → signups → customers new customer
│ │
▼ ▼
end does something that reaches others
(refers, shares, publishes, invites)
│
▼
some of them become customers
│
└──────→ back to the top
A funnel needs fresh input every cycle. A loop feeds part of its own input back.
The common loops
| Loop | The output that feeds back | Example shape |
|---|---|---|
| Referral | A customer invites someone | Give £10, get £10 — Referral Programmes |
| Viral / product | Using the product exposes it to non-users | A shared document, an invoice with “sent with…” in the footer |
| User-generated content | Users create pages that rank or get shared | Reviews, Q&A, public profiles indexed by search |
| Company content | Content brings visitors, some convert, revenue funds more content | An SEO programme — Search Engine Optimisation |
| Paid | Customers generate margin that’s reinvested in ads | Contribution from cohort 1 funds acquisition of cohort 2 |
The paid loop is the one most ecommerce businesses actually run, and its speed is set by Payback Period: the faster margin comes back, the faster it can be respent.
The multiplier
k (the viral coefficient) is the number of new customers each customer brings in: invitations sent per customer × the share of invitations that convert.
Worked example. Start with 100 customers from paid acquisition.
k = 0.3 (each customer brings 0.3 new ones)
cycle new from loop running total
0 100 (paid) 100
1 100 × 0.3 = 30 130
2 30 × 0.3 = 9 139
3 9 × 0.3 = 2.7 141.7
4 2.7 × 0.3 = 0.8 142.5
... → converges on 100 ÷ (1 − 0.3) = 142.9
k = 0.8
cycle 1 80 cycle 2 64 cycle 3 51 cycle 4 41 ...
→ converges on 100 ÷ (1 − 0.8) = 500
The formula: total customers = paid customers ÷ (1 − k), for any k below 1.
In plain terms: below 1, a loop doesn’t grow on its own — it multiplies whatever you put in. A k of 0.3 makes every paid customer worth 1.43; a k of 0.8 makes each one worth 5. Only at k ≥ 1 does it grow without new input, and almost nothing sustains that.
Why this is counterintuitive. “Viral” gets used for anything with a share button. Most products that describe themselves as viral have a k well below 0.5. That’s still valuable — it’s a 1.4× discount on acquisition cost — but it’s a discount, not an engine.
Cycle time matters as much as k. A loop with k = 0.5 that turns over weekly beats one with k = 0.7 that turns over yearly for any horizon a business plans on.
Where they fail
- The loop doesn’t close. Customers are asked to refer, few do, and the ones who do bring people who don’t buy. Every step in the loop is its own conversion rate, and a loop is only as strong as their product
- Counting loop customers who’d have come anyway. Referred customers are often people already aware of you. Incrementality applies here as much as to paid media — Incrementality Testing
- The loop degrades as the market saturates. The first customers invite people who haven’t heard of you; later ones invite people who already have
- Fraud. Any loop with a cash incentive attracts self-referral — Referral Programmes
Instrumenting one
- Measure each step, not just the outcome. Share rate, invite acceptance, invitee conversion — k is their product, and the weak step is where to work
- Track the generation of each customer (paid, loop-generation 1, 2…) so the multiplier is measured rather than assumed — Cohort Analysis
- Watch cycle time alongside k