What a growth loop is
The loop framing is an alternative to the funnel for describing how products that grow mostly through their own use actually grow. The funnel's picture is a pipe: marketing fills the top, some fraction reaches the bottom, and next period the filling starts again. The loop's picture is a circle: the people, content, data or money that come out of one cycle are what fills the next, and marketing's job is to start and widen the circle rather than to keep pouring.
The framework forces three questions. What is the input: new users, a budget, a piece of content, an invitation, a review? What is the action that turns the input into something: using the product, sharing, publishing, buying, referring? And what is the output, and how much of it comes back as input? That last clause is the heart of the model. A loop where each cycle returns more than it started with compounds; one where it returns less is a funnel wearing a circle, and still needs an external source to keep going.
What the loop makes visible is dependence. A business that believes it grows by word of mouth can draw the loop and discover that the share returned per cycle is small and the cycle is slow, and that the growth it sees comes from paid acquisition it was not counting. What the loop hides is the detail inside each arrow: the marketing funnel framework still measures how many of the people entering a step reach the next, and the loop only tells you where they came from and where they go. The growth marketing hub places loops alongside economics and experimentation as the three parts of a growth system.
Loop types
| Loop type | Input | Action | Output that returns | What it hides |
|---|---|---|---|---|
| Word of mouth and viral | New user | Uses the product and invites or shows others, because the product is better with them | Some invitees become new users | Depends on a product that is visibly better shared; decays as the easy invitees are used up |
| User-generated content | New user or contributor | Creates content the product hosts: reviews, listings, profiles, answers | Content is found by search or social and brings new users or contributors | Quality and freshness of content; search rules can change the return without warning |
| Company content | Budget or editorial effort | Publishes content that ranks or is shared | Visitors convert, and revenue or audience funds more content | Slow cycle; attributing revenue back to content is hard |
| Paid | Budget | Buys attention that converts to customers | Margin from those customers is reinvested in the next cycle's budget | Entirely dependent on the gap between cost per customer and margin, and on rising media costs |
| Sales | Leads or accounts | Closes, delivers and expands accounts | Referrals, case studies and expansion revenue fund more sales capacity | Long cycles; a few large accounts can dominate the figures |
| Marketplace or network | Supply or demand on one side | Attracts the other side, which attracts more of the first | Each side's growth is the input for the other | Cold-start problem; local density matters more than totals |
| Referral programme | Existing customer | Refers for an incentive | Referred customers, some of whom refer in turn | Incentive cost and fraud; often a small loop presented as a large one |
Most businesses run several of these at once and one of them dominates. A useful first exercise is to draw every loop the business claims to have and then try to put a number on the return of each; the loops that survive the attempt are the real ones. The AARRR framework is a helpful companion here, because its referral and revenue stages are precisely the outputs that feed loops, and a business that measures those stages already has the figures to size its loops. The growth loops in practice guide goes into how each type is built and where it leaks; this page stays with the framework. For a referral loop, the referral reward test designs the comparison of reward structures that tells a business which version of the incentive is worth its cost.
Mapping a loop
A loop map is a short cycle of boxes and arrows with a number on each arrow. Start with the input and name it precisely: not "users" but "users who complete onboarding", not "content" but "guides that reach a ranking position". Then the action, stated as behaviour that can be observed: creates a shared project, posts a review, publishes a guide, spends a budget. Then the output, in the same unit as the input where possible, so that the return per cycle can be read directly. Then the arrow back to the start, with the share of the output that becomes input.
Every arrow has a conversion attached. The share of users who take the action is a CVRShare of visits or clicks that convert. in the loop's own terms; the share of outputs that become inputs is another. The product of the rates around the loop is the loop's return per cycle, and it is the single number the map exists to produce. Beside it goes the cycle time: how long one trip around takes, which decides how many cycles a year the loop can run and therefore how much compounding is available.
The map also records leaks, meaning the points where the loop loses what it could have kept. Invitees who never open the invitation, content that never ranks, referred customers who churn before referring. Each leak is a candidate for work, and the map shows which leaks sit on high-volume arrows. Map the current state from data, not the intended state from the product plan, and date the map. Loops change as products and channels change, and a map of last year's loop is a story.
Loop metrics
The loop's primary metric is the return per cycle: the output that comes back as input, divided by the input that started the cycle. A return above one means the loop grows on its own; below one it needs external input to sustain, and the shortfall is what marketing has to supply. The second metric is cycle time, because a modest return that cycles quickly can outgrow a strong return that cycles slowly. The third is decay: how the return changes as the loop runs, since most loops weaken as the easiest inputs are exhausted. How these three combine into the customers a loop actually adds, generation by generation and cycle by cycle, is set out in how a growth loop compounds.
Beneath these sit the arrow-level conversion rates, which locate problems the way adjacent funnel rates do. Treat them exactly as funnel rates: volume alongside, segmented by cohort, with a written definition of what counts. Cohort views matter more for loops than for funnels, because a loop's return is only visible when the people who entered in a period are followed long enough to see what they bring back.
Finally, a loop is only worth running if what it compounds is worth having. CACCost to acquire one new customer. for the customers a loop produces, compared with the LTVTotal value a customer brings over time. of those customers, decides whether the loop should be fed or starved, and a paid loop in particular lives or dies on that comparison. A viral loop that brings in users who never pay compounds a cost. The loop framework shows how growth happens; unit economics says whether it should. For a marketplace loop, where each side is the input for the other, the guide to measurement for marketplaces shows how to measure each side and the liquidity that connects them, and the guide to growing a two-sided marketplace takes the operator's decisions in order, from finding the side that limits matches to expanding area by area.
Worked loop (illustrative)
A project-management tool sizes its sharing loop
A project-management tool for small agencies believes it grows through shared boards and maps the loop to check. Input: 1,000 new sign-ups in a month. Action: 35% of them create a board and invite at least one collaborator, sending an average of 2.4 invitations each, so 840 invitations go out. Output: 22% of invitees sign up, producing 185 new sign-ups. Return per cycle is therefore 0.185, with a median cycle time of 11 days, and the loop decays: cohorts followed for 3 cycles show the invitation rate falling from 35% to 19% in the second cycle as later sign-ups are smaller agencies with fewer colleagues to invite. The loop is real but far from self-sustaining; the 1,000 monthly sign-ups come overwhelmingly from paid search and a content programme, and the sharing loop adds roughly a fifth on top. The leak analysis shows that 41% of invitations are never opened because they arrive as a bare email with no preview of the board. A redesigned invitation with a live preview lifts invitee sign-up to 31%, taking the return per cycle to 0.26 without changing the number of invitations. The team records the loop with a date, keeps the paid and content loops as the primary engines, and sets a quarterly check on the return figure.
Limits
The loop framework is a model of where growth comes from, not a guarantee that it compounds. Most loops have a return per cycle well below one, and the honest reading of a loop map is usually that the business still depends on external input, which is a useful finding but not the exciting one teams hope for. Loops also decay: the first cohorts invite more, the first content ranks more easily, and a return measured early flatters the long run. A loop map without cohort tracking over time is a snapshot presented as a trend.
The model is also quiet about brand, pricing, and anything that does not fit a circle. A business can grow for years on a strong brand and sensible pricing without any loop that would survive the mapping exercise, and forcing such a business into loop language produces diagrams rather than insight. For businesses with long sales cycles, a loop may take so long to turn that the metrics are unmeasurable within a planning period.
Finally, loops are often compared with the flywheel framework, and the two are easily confused. The flywheel describes momentum and friction in a cycle of attract, engage and delight, and is a way of thinking about where effort reduces resistance. The loop framework is narrower and more quantitative: a specific input, a specific action, a specific return. The flywheel is a metaphor for a whole business; a loop is an accountable mechanism within it, and a business usually has several.
For mobile products, the guide to growth loops built to drive app installs adds the deep link, the store page and a ranking loop as joints the web version never has. For a new product with no users yet, growth loops for a product launch covers seeding a connected first group, running a referral waitlist and reading the loop by launch cohort.
Common mistakes
- Drawing the intended loop. Map the current state from data; a loop from the product roadmap is a plan, not a measurement.
- No number on the return arrow. A loop without a return per cycle is a diagram. The share that comes back is the whole framework.
- Mistaking a small loop for the engine. A referral programmeA scheme that rewards existing customers, and often the friends they invite, when a recommendation brings in a new customer. with a return far below one is a supplement to acquisition, not a replacement.
- Ignoring cycle time. Two loops with the same return grow at very different rates if one cycles weekly and the other quarterly.
- Measuring once. Loops decay; the return has to be tracked by cohort over time, not read from the first month.
- Compounding the wrong thing. A loop that multiplies users who never pay compounds a cost. Check the economics before feeding it.
