Growth loop

A growth loop is a reinforcing cycle: an action by a user produces an output (content, a referral, a link, an integration) that pulls a new user in, who then produces more of the same output. Pinterest is loop: new users save pins → pins rank in search → searchers discover Pinterest → some sign up and save pins. The loop is closed; it does not run out as long as users keep doing the action.

Loops differ from funnels in a structural way. A funnel is one-shot: traffic arrives, some converts, the rest is lost forever. A loop is compounding: every retained user is also an acquisition channel for the next cohort. The implication for experiments is huge — a test that improves a loop step (e.g., the rate at which new users invite teammates) compounds across cohorts in a way a funnel-step test cannot. Pick experiments that move the loop is bottleneck, not the funnel is widest spot.

When to use it

Identify your loops before designing experiments. A test that moves a loop-step compounds; the same-sized test on a funnel-step does not. Diagnose where the loop leaks and target experiments at that step.

What this looks like in practice

There are roughly four canonical loop types: viral (users share, recipients sign up), content (users create content that ranks in search or social, drawing more users), paid (revenue from users funds ads to acquire more users), and sales-led (customers refer or expand into adjacent teams). Most products have one dominant loop and one or two weaker ones. Knowing which is dominant tells you where to spend experiment cycles.

Loops have step-level conversion rates that multiply. If new users invite at 30%, invitees accept at 50%, and accepted invitees become inviters at 60%, the loop coefficient is 0.30 × 0.50 × 0.60 = 0.09 — every 100 new users produce nine more in the next cycle. Move any single rate by a meaningful amount and the cumulative effect over many cycles is dramatic. The math rewards loop-step optimization more than funnel-step optimization for products that already have a loop.

Not every product has a loop, and pretending it does is expensive. If your customers do not naturally produce content, refer others, or generate links and integrations as a side effect of using the product, you are running a funnel and should design experiments accordingly. Founders who chase "viral coefficient" experiments on fundamentally non-viral products waste quarters. Diagnose the structure first; pick the experiments second.

A worked example

Common mistakes

• Calling everything a loop.A loop requires that user output produces input for the next user. "We send emails to existing customers" is retention marketing, not a loop. The discipline is in the structural definition, not the vibe.
• Optimizing the wrong loop step.If invite-acceptance is already 90%, lifting it to 95% does almost nothing. Find the lowest-conversion step in the loop and target that — the math is dominated by the bottleneck.
• Ignoring loop quality for loop volume.A 10x lift in invites that brings in low-quality users who never invite anyone themselves shrinks the loop coefficient. Verdicts on loop-step experiments must check the next cycle, not just the immediate output.

Related terms