Not a funnel, not a buzzword. A growth loop is a system where the output of one cycle becomes the input for the next. Here is the real definition, named examples, and one popular stat you should stop repeating.
Written for founders and growth teams who keep hearing "growth loop" in pitch decks and want the concrete, sourced version rather than the buzzword.
A growth loop is a self-reinforcing system where the output of one cycle feeds directly back in as the input for the next cycle, a referral produces a new user who refers again, a piece of content attracts a visitor who creates more content. This is different from a linear marketing funnel, where more input at the top produces more conversions at the bottom, and growth stops the moment you stop adding to the top.
The framing was popularized by Brian Balfour and Reforge in the 2019 post "Growth Loops are the New Funnels," and it has since become a standard way growth teams describe compounding acquisition mechanisms, Dropbox's referral program and TikTok's share-and-remix behavior are the two most commonly cited real examples.
A growth loop is a closed system, not a straight line. Instead of traffic entering at the top and exiting as a conversion at the bottom, a loop takes what comes out of one cycle and routes it back in as fuel for the next cycle. That reinvestment step is the whole idea, without it, you just have a repeating tactic, not a loop.
Brian Balfour, former VP of Growth at HubSpot and founder of Reforge, laid out the distinction in a 2019 Reforge post titled "Growth Loops are the New Funnels." His argument was that the fastest-growing products are better modeled as systems of loops than as a single funnel, because loops compound over time while funnels only scale with added spend.
In one sentence
A growth loop is a cycle where today's output becomes tomorrow's input, so the system can keep growing itself instead of needing a fresh push from the outside every time.
Funnels are not wrong, they are just a narrower lens. A funnel is useful for describing one stage of a system, but it cannot explain how a product keeps growing without added spend. A loop can.
| Dimension | Funnel | Loop |
|---|---|---|
| Direction | One-way, top to bottom | Circular, output feeds back into input |
| Growth shape | Linear, roughly proportional to spend | Compounding, each cycle can produce more than the last |
| What happens at the bottom | A conversion, then the system stops | The conversion becomes a new input that restarts the cycle |
| Primary lever | Spend more to get more at the top | Improve the conversion rate at each stage of the cycle |
| Failure mode | Growth stalls the moment spend stops | Growth stalls only if a stage in the cycle breaks |
| Best framed by | Brian Balfour and Reforge, "Growth Loops are the New Funnels," 2019 | Same source, describing loops as the more accurate model for compounding growth |
Whatever type of loop you are looking at, it breaks down into the same four parts, and the loop only works if all four hold together.
The starting resource that enters the loop, a new visitor, an existing user, a piece of content, or a unit of ad spend.
What the input does inside the product, an action that is valuable to the user for its own sake, not just something they do to please the growth team.
What the action produces, an invite sent, a piece of content published, a review left, a referral link shared.
The mechanism that feeds the output back in as a new input, the invite gets clicked, the content gets discovered, the referral gets redeemed, restarting the cycle.
An existing user invites a new user, often with an incentive on one or both sides, and the new user becomes a fresh input who can invite more people. Dropbox's give-space-get-space program is the canonical named example.
A user creates content, that content attracts new visitors through search or social feeds, and a share of those visitors become users who create more content. Pinterest's pin-and-repin behavior is a well documented example.
Revenue from converted users is reinvested into paid acquisition for more users, and if payback is fast enough, the reinvestment compounds instead of just repeating flat spend.
Usage or data from existing customers, logos, case studies, referrals, generates new sales conversations, which produce more customers who feed the same loop.
Creators publish on a platform, the platform distributes that content to attract an audience, and the audience includes future creators. TikTok's share-and-remix behavior is the most cited modern example.
A Dropbox user who referred a friend got extra storage, and the referred friend got storage too. Once that friend signed up, they had the same incentive to refer their own friends, restarting the exact same cycle. Signups reportedly grew from about 100,000 to 4,000,000 over 15 months during this period, a figure widely cited in growth writeups covering the program. The loop worked because the reward was the product itself, storage, not an unrelated discount or gift card.
According to a First Round Review interview with Pinterest's former growth lead, Pinterest ran two related content loops: one where user boards and repins were made searchable and indexed by Google, and another where a user saving an idea led a friend to discover it and create their own pins from it. Both loops turned everyday product usage into new organic discovery, not a bolted-on referral prompt.
A user posts a video, the platform's recommendation feed distributes it to a wide audience regardless of the poster's follower count, some viewers who are not yet on the platform join to remix or respond to the trend, and those new users post their own videos, restarting the cycle. This distribution mechanic is commonly cited as one of the clearest modern examples of a content-driven growth loop.
Loops are not the only way to get found, either. Tools like MediaFast help founders show up directly in the Reddit threads where their exact audience is already asking for a solution, which works alongside a loop rather than replacing one.
While your growth loop builds momentum, MediaFast finds the live Reddit conversations where your exact audience is already asking for a product like yours.
A specific claim circulates widely in growth-loop content: that growth loops "cut CAC by 40%." It shows up in blog posts, decks, and social threads, usually stated flatly as fact and rarely attached to a named study, company, or research firm.
We could not find a credible, dated primary source for this exact figure. That does not mean loops never lower acquisition cost, a strong referral or content loop plausibly can, but a specific "40%" number with no named source behind it should be treated as an example of the kind of unverified statistic that gets repeated in marketing content until it sounds like established fact. If you see this number cited elsewhere, ask for the underlying source before repeating it yourself.
This page deliberately does not restate that "40%" figure as fact, and does not use other unsourced round numbers, like a specific dollar CAC or content-cost comparison, that circulate in the same genre of content without traceable sources.
Six steps, roughly in order, from finding an existing loop to sustaining one over time.
Map your product's existing loop, if one already exists informally
Most products already have a weak, unintentional loop, happy customers occasionally refer others. The first job is to find it and measure its current strength before trying to engineer a new one.
Pick the loop type that matches how your product is actually used
A B2B tool with no natural sharing motion is a poor fit for a viral loop, but might have a strong content or sales loop instead. Force-fitting the wrong loop type wastes engineering effort.
Instrument every stage of the cycle, not just the final conversion
A loop is only as strong as its weakest stage. Without stage-by-stage measurement, a team cannot tell whether a slowdown is at the invite step, the click step, or the signup step.
Improve the conversion rate at each stage before adding volume
Because loops compound, a small improvement in one stage's conversion rate compounds across every future cycle, which is usually a better investment than simply pushing more input into a weak loop.
Set a cycle time and track it explicitly
Cycle time, how long it takes for one output to become the next input, determines how fast a loop compounds. A loop with a one-day cycle time compounds far faster than one with a one-month cycle time, even at the same conversion rate.
Watch for decay and keep reinvesting in the mechanism
Loops weaken over time as channels saturate or novelty fades. Reforge's framing treats loops as systems that need ongoing investment, not a one-time mechanism you build once and forget.
Building a loop is one part of a growth system. The how many marketing channels should a startup focus on guide covers how to prioritize a loop against other channels rather than chasing every one at once.
Not every product should chase a loop before it has the fundamentals in place.
A design tool where users share finished work, a scheduling tool where invitees see the brand, these have a loop to strengthen rather than invent from nothing.
Loops are optimized stage by stage. Without enough volume to see where a cycle breaks down, you are guessing rather than measuring.
As with Dropbox's storage reward, the strongest loops tie the incentive directly to the value the product already delivers, rather than an unrelated bonus.
The product has no natural sharing, content, or referral behavior to build on, forcing a mechanic onto it usually feels bolted-on
You do not yet have enough users or usage to measure and improve stage-level conversion rates
You have not validated product-market fit yet, a loop compounds whatever is already working, it does not fix a product people do not want
Each of these turns a promising loop into a mechanic that quietly stalls.
A newsletter that runs every week is a routine, not a loop, unless its output genuinely becomes a new input. The word gets applied loosely to anything cyclical, which drains it of meaning.
Dropbox's loop worked because the underlying product, more storage, was something people already wanted. A referral prompt bolted onto a mediocre product produces a weak loop no matter how well engineered the mechanics are.
Without stage-level data, a team cannot tell whether the loop is decaying because fewer people are sharing, fewer shares are getting clicked, or fewer clicks are converting.
Two loops with identical conversion rates but different cycle times compound at very different speeds. A slow loop can look healthy on a monthly report while quietly underperforming a faster one.
Growth loops get pitched internally with round, dramatic statistics that have no traceable source, which sets unrealistic expectations and makes it harder to evaluate whether a specific loop is actually working. See the callout below.
Viral loops are one type among several. A content loop, a sales loop, or a paid loop can be the right fit, and forcing a viral mechanic onto a product without natural sharing behavior usually just annoys users.
Viral or referral loops are the most talked-about type, but content loops, paid loops, and sales loops are all equally valid loop structures. Reforge's original framing covers all of them, not just referrals.
Loops decay. Channels saturate, novelty fades, and platforms change their distribution rules. A loop needs ongoing measurement and reinvestment, not a one-time build.
A loop is made up of stages, and each stage still behaves like a mini funnel with a conversion rate. The funnel view did not disappear, it became a component inside the larger loop view rather than the whole picture.
A closed system where the output of one cycle becomes the input for the next, producing compounding rather than linear growth, as framed by Brian Balfour and Reforge in 2019.
A one-directional model of growth, more input at the top produces more output at the bottom, with no built-in mechanism to reinvest that output back into the top.
How long it takes for one output of a loop to become a new input, a key variable in how fast a loop compounds.
The average number of new users each existing user brings in through a referral or invite loop, a coefficient above 1 means the loop grows on its own without added input.
Growth where each period's output becomes part of the next period's input, producing acceleration over time rather than a flat, repeating rate.
A loop where user-created content attracts new users through search or social discovery, and a share of those new users create more content.
The natural weakening of a loop over time as a channel saturates, novelty fades, or a platform changes its distribution rules, requiring ongoing reinvestment to sustain.
The primary sources behind the definitions and examples on this page.
Reforge (Brian Balfour)
The original 2019 framing that popularized growth loops as a replacement for the funnel model of growth.
First Round Review
A first-hand account of how Pinterest engineered two distinct content loops around its pin-and-repin behavior.
ReferralRock
A detailed writeup of Dropbox's give-space-get-space referral loop and its widely cited signup growth over 15 months.
A growth loop is a system where the output of one cycle feeds the next, which is why it compounds while a funnel only scales with added spend. Dropbox's referral program and Pinterest and TikTok's content loops are the real, named examples worth studying, not the vague "loops beat funnels" framing that circulates without any grounding.
Be skeptical of round, unsourced numbers attached to growth loops, the "40% CAC reduction" figure being the clearest example. Build a loop because your product already has the underlying motion for one, then measure it stage by stage rather than repeating a stat you cannot trace.
A growth loop is one piece of a larger growth system, these guides cover the rest.
The questions founders and growth teams ask most when a growth loop gets pitched to them.
A growth loop is a closed, self-reinforcing system where the output of one cycle becomes the input for the next cycle. A user's action produces something, an invite, a piece of content, a referral, that brings in a new user who repeats the same action, so the loop keeps feeding itself instead of needing fresh input from the outside every time.
A funnel is one-directional, more traffic at the top produces more conversions at the bottom, and growth stops the moment you stop feeding the top. A growth loop is circular, the output of a cycle becomes a new input, so growth can compound instead of staying flat and proportional to spend. Brian Balfour and Reforge popularized this distinction in their 2019 post "Growth Loops are the New Funnels."
Dropbox's referral program is a commonly cited example: a user gets extra storage for referring a friend, the friend signs up and gets storage too, and the friend then has the same incentive to refer their own friends, restarting the cycle. Pinterest and TikTok are cited examples of content-based loops, where user-created content attracts new users who then create more content.
That specific "40% CAC reduction" figure circulates widely in marketing content but does not trace back to a credible, dated primary source, no named study, company report, or research firm attaches to it consistently. Treat it as an unverified number rather than a fact, and look for loop performance data from a named source before repeating a specific percentage.
The commonly described types are viral or referral loops, where users invite other users, content loops, where user-generated content attracts new users through search or social discovery, paid loops, where converted revenue is reinvested into acquisition, and sales loops, where existing customer activity generates new sales conversations.
No. Not every product has a natural sharing, content, or reinvestment motion, and forcing a loop mechanic onto a product that lacks one usually produces a weak result. Many early-stage products are still better served by a well-run funnel and direct distribution work before a loop becomes the right investment.
