| url | https://www.reforge.com/blog/growth-loops |
|---|---|
| raw | raw/highlights-growth-loops.json |
TL;DR: The fastest-growing products are systems of loops, not funnels. A loop’s output reinvests as its input, creating compounding growth that funnels structurally cannot match. The funnel model — acquisition → activation → retention → revenue → referral — is wrong, and almost everyone is still using it.
What it means
Balfour argues that the funnel model creates silos that optimize against each other. Marketing brings in low-quality users to hit top-of-funnel targets, which tanks retention. Growth and product end up fighting because the funnel encourages them to. Loops force you to answer a different question: “How does one cohort of users lead to another cohort?” This single question reframes growth from a linear pipeline into a compounding system, and the difference is enormous over time.
Not all loops are equal. The fastest-growing products are powered by 1–2 major loops, not dozens of weak ones. Pinterest’s core loop: user signs up → saves content → Pinterest distributes that content to search engines → new users find it via Google → sign up. Each cycle feeds the next. The loop produces its own users. That’s the key feature — loops create their inputs from their own outputs, while funnels consume inputs you have to buy from somewhere else.
The argument
Loops combine product, channel, and monetization. Because loops are specific to your product and your users, they’re harder to replicate than generic tactics. Strategies not specific to your product “by definition can be replicated with ease” — and copied tactics trend toward zero effectiveness within months (moats). The loop is unique to your business; the tactic isn’t.
Loops are the mechanism behind network-effects. A network effect is a specific kind of loop: more users → more value → more users. But Balfour’s framework is broader — content loops, data loops, and economic loops all compound without requiring network density. ChatGPT’s data loop (cheating-is-all-you-need) is a textbook example: more users → more conversation data → better model → more users. No social graph required (distribution).