Growth That Runs Downhill
There is a question that separates products that grow easily from products that grow only as hard as you push them: when a new user joins, does anything about their joining make the next user more likely to join? In most products the answer is no. Each user is acquired independently, at a cost, through a channel you have to keep feeding. In a few products the answer is yes, and those products grow in a fundamentally different way. They grow themselves.
The difference is the difference between a funnel and a loop. A funnel is a pipe you pour acquisition into; what comes out the bottom is growth, and the moment you stop pouring, the output stops. A loop is a system where the output of growth feeds back into the input, so that growth produces more growth. The first scales with your spending. The second compounds.
This essay is about why loops beat funnels, what the common loop types look like, the reinvestment mechanic that makes a loop a loop, how to design one deliberately, and how to measure whether yours is actually working. The aim is to change how you think about growth: not as a pipeline to optimize, but as a machine to build.
The Funnel and Its Ceiling
The funnel is the default mental model of growth, and for good reason. It is simple and it describes something real. Users enter at the top through awareness, move down through consideration and signup and activation, and a fraction emerge at the bottom as retained, paying customers. Each stage has a conversion rate. Optimize the rates, widen the top, and you get more out the bottom. Generations of growth teams have been built on exactly this.
The funnel is not wrong. The problem is that it is linear, and linear systems have a ceiling. The output is a function of the input. Double your output and you must roughly double your input: twice the ad spend, twice the sales headcount, twice the content. There is no leverage. The funnel does not remember its own success. Yesterday's converted users do nothing to fill the top of the funnel today.
The Hidden Cost of Funnel Growth
A funnel has a second, harsher property: its inputs tend to get more expensive over time. The cheapest, most obvious audience is the first you reach. As you exhaust it, you pay more to reach the next-best audience, and more again for the one after that. Channels saturate. Auction prices rise as competitors crowd in. A funnel-only growth model is therefore not just linear but linear with rising costs, which means your efficiency erodes precisely as you try to scale. You run faster to stay in place.
What a Growth Loop Is
A growth loop is a closed system in which the output of one cycle becomes the input to the next. New users do something that, as a byproduct of getting value, brings in more new users, who do the same. The defining feature is the feedback edge: the arrow that runs from the bottom of the process back to the top. A funnel is a line. A loop is a circle.
Consider the canonical example. A user signs up for a file-sharing product, stores files, and shares some of them with collaborators. To open a shared file, those collaborators sign up. Some of them store and share their own files, bringing in more people. The act of getting value, sharing a file, is the same act that recruits the next user. No separate acquisition campaign created that growth. The product's normal use generated it.
The Loop Is Built Into the Product, Not Bolted On
This is the part that trips teams up. A loop is not a referral program you add at the end. The strongest loops are intrinsic: the user grows the product by doing the thing they came to do. When sharing a document recruits a collaborator, growth is a side effect of value. When you have to bribe users to invite friends to a product whose normal use involves no one else, you are bolting a loop onto a product that does not naturally have one, and it shows. Bolted-on loops are weak, expensive, and easy to game. Intrinsic loops are durable because they are powered by the same engine as the product's core value.
Types of Loops
Loops come in several shapes. Most products that compound run one dominant loop, sometimes two. The common families:
Viral Loops
A user, in the course of using the product, exposes a non-user to it, and some fraction of those non-users convert. The file-sharing example is a viral loop. So is a messaging app where you message someone who has to install it to reply. The fuel is direct user-to-user contact. Viral loops can be fast, because the cycle time, from a user joining to that user recruiting the next, can be measured in days or hours.
Content Loops
Users create content that gets indexed and discovered by strangers, who arrive, find value, and create more content. A question-and-answer site where users post questions that rank in search, drawing in new visitors who ask their own questions, runs a content loop. A reviews site where each review becomes a page that attracts searchers works the same way. The fuel here is not direct contact but discoverability: the content users produce is the acquisition channel.
Paid Loops
A paid loop sounds like a contradiction but is one of the most powerful. Users generate revenue; a portion of that revenue is reinvested into paid acquisition; the new users generate more revenue; more gets reinvested. The loop is closed not through the product's social mechanics but through economics. As long as the revenue a user generates exceeds the cost to acquire the next one within an acceptable payback period, the loop spins, and it can spin at enormous scale because money is a fungible, unlimited input in a way that user-to-user contact is not.
The catch with paid loops is that they live or die on unit economics. The loop only sustains if lifetime value comfortably exceeds acquisition cost, and it only compounds if you actually reinvest the margin rather than banking it. Plenty of companies have a paid loop that works at small scale and breaks when channel costs rise as they spend more. The economics, not the mechanic, are the whole game.
The Reinvestment Mechanic
Here is the heart of why loops compound and funnels do not. A loop has a reinvestment step: some output of the cycle is fed back as input. In a paid loop, the reinvested quantity is money. In a viral loop, it is user-generated invitations. In a content loop, it is user-generated pages. Whatever the currency, the principle is identical: the product takes a share of what this cycle produced and uses it to seed the next cycle.
This is exactly the structure of compound interest. In a funnel, your input each period comes from outside, from your budget or your effort. In a loop, your input each period comes partly from the previous period's output. That is what creates the exponential rather than linear shape. The more the previous cycle produced, the more is available to invest in the next, the more the next produces, and so on. The loop borrows its own success forward.
Reinvestment Has to Be Deliberate
The reinvestment step is the one most often left implicit, and that is a mistake. If you do not consciously decide how much of each cycle's output feeds the next, you tend to leak it. In a paid loop, the leak is spending the margin on something other than acquisition. In a content loop, the leak is letting user content go unindexed or undiscoverable, so the pages produce no new visitors. Designing a loop means designing the reinvestment explicitly: what gets fed back, how much, and through what path.
Designing a Loop
A loop is not luck. You can design one, though you cannot conjure one where the product has no natural mechanism. Designing a loop means making each edge of the cycle explicit and then asking whether the cycle actually closes.
Map the Cycle as Steps
Write the loop down as a sequence that returns to its start. A clean viral loop might read: a new user signs up, the user creates something worth sharing, the user shares it with non-users, some non-users click through, some of those sign up. The last step has to connect back to the first, or you do not have a loop, you have a funnel with extra stages. The discipline of writing it as a closed sequence forces you to confront whether the feedback edge truly exists.
Find the Weakest Edge
Every edge in the loop is a conversion: the share of users who advance to the next step. The loop's strength is governed by the product of these conversions, which means the weakest edge dominates. If users love the product and share enthusiastically but the shared link lands on a confusing signup page where most visitors bounce, that one edge throttles the entire loop. Designing a loop is largely the work of finding and widening its weakest edge, then finding the next weakest.
Shorten the Cycle Time
Loop strength is not only about conversion; it is about speed. A loop where the cycle takes a day compounds far faster than an identical loop where it takes a month, because it spins more times in the same period. When you design a loop, ask what is slowing each cycle and whether you can compress it. Reducing the time between a user joining and that user recruiting the next is often as valuable as improving the conversion rates themselves.
Measuring Loop Strength
A loop you cannot measure is a story you are telling yourself. The measurement question is simple to state: for each user who enters the loop, how many new users does the loop eventually produce, and how fast? Two quantities matter.
- The amplification per cycle. How many new users does each user in this cycle generate for the next? In a viral loop this is the familiar branching factor: above a critical threshold the loop grows on its own; below it, the loop decays and merely amplifies whatever you put in from outside.
- The cycle time. How long does one full turn of the loop take? Two loops with the same amplification but different cycle times produce wildly different growth, because the faster one compounds more often in the same window.
Most loops, honestly measured, sit below the self-sustaining threshold. That is not a verdict of failure. A loop with amplification below the critical point still helps enormously: it multiplies your other acquisition. Every user you bring in through paid or content channels also flows through the loop and produces some additional users on top. A sub-critical loop is a multiplier on the rest of your growth, even when it cannot sustain growth alone.
Why Loops Are Durable
The final reason to prefer loops is durability. Funnel growth is rented. You hold your position only as long as you keep paying, and the rent rises. The moment you stop feeding a channel, that channel's contribution decays. Loop growth, when it is intrinsic, is owned. The growth mechanism is built into how the product delivers value, so it keeps running as long as the product is used.
Loops also tend to deepen the product's defensibility. A content loop produces a growing library of pages that a competitor cannot replicate quickly. A viral loop seeds a network whose value to each user rises as more users join, which raises the cost of leaving. The same mechanism that grows the product also entrenches it. Funnels rarely produce this; optimizing a signup flow does not make the product harder to leave. Loops, by knitting growth into value, tend to build a moat as a side effect.
The Honest Limit
None of this means funnels are obsolete or that every product can have a loop. Many genuinely cannot; their value is delivered to a solitary user with no natural reason to bring in others, and bolting on a referral scheme will not change that. For those products, a well-run funnel with disciplined paid economics is the right answer, and pretending otherwise wastes years chasing a loop that was never there. The skill is honesty about which kind of product you have. Build the loop where the product affords one, and run the funnel well where it does not. The mistake is not running a funnel. The mistake is running a funnel while believing it will someday compound on its own.
A Final Word
The shift from funnel thinking to loop thinking is one of the more consequential reframes in product growth. A funnel asks "how do I move more users through the pipe?" A loop asks "how do I make the product grow itself?" The first question has a ceiling set by your budget and your effort. The second, when the product affords it, has no ceiling, only a rate.
Look at your own product and find the feedback edge, if there is one. Trace whether the act of getting value also, as a byproduct, brings in the next user. If it does, you have a loop worth strengthening, and your job is to widen its weakest edge and shorten its cycle. If it does not, be honest about that and run your funnel with discipline. Either way, you will have stopped confusing a pipe with a machine, and that clarity alone will change where you spend your effort.
Key Takeaways
- A funnel is linear and rented: output scales with input, inputs get more expensive over time, and growth stops when you stop paying. A loop feeds its own output back as input, so growth compounds.
- The strongest loops are intrinsic, where users grow the product by doing the thing they came to do. Bolted-on referral schemes are weak substitutes for a loop the product naturally affords.
- Loops come in families, viral, content, and paid, but each depends on a deliberate reinvestment step that feeds part of this cycle's output into the next. Leak the reinvestment and the loop stops compounding.
- Loop strength is amplification per cycle times speed of cycle. The weakest conversion edge dominates, and even a sub-threshold loop is valuable because it multiplies all your other acquisition.
- Loop growth is durable and tends to build a moat, because the mechanism that grows the product is built into how it delivers value. But be honest: not every product affords a loop, and a well-run funnel is the right answer when it does not.