Growth Loops
Growth loops are self-reinforcing cycles where each user's behaviour helps acquire or activate the next. Here's how to model them, measure loop factor, and decide which loop to amplify.
Growth Loops
A self-reinforcing acquisition cycle where one user's behaviour creates the conditions for the next user's acquisition or activation.
A growth loop is a closed system: a user takes an action, that action produces an output (a referral, a piece of UGC, a review, a re-targeting signal, revenue reinvested in ads), and that output feeds the input of the next cycle. The same loop runs again, but with a slightly larger base of users each time.
Loops are the strategic frame above individual experiments. A funnel asks 'how do we move this cohort from step A to step B?' A loop asks 'what does each converted customer produce that brings the next customer in?' On a healthy loop, paid acquisition becomes a starter, not a permanent cost — compounding does the work.
The mental model matters because it changes what you optimise. A funnel mindset pushes you to fix conversion-rate leaks one step at a time. A loop mindset pushes you to ask which output of a converted buyer — a review, a referral code, a shareable unboxing — is most likely to recruit the next buyer, and how you compress the time between those events.
Most online stores run several loops at once, with very different strengths. A paid loop reinvests gross margin into ads. A content loop turns customer questions into indexed product pages. A UGC loop turns post-purchase moments into TikToks and reviews that warm up cold traffic. Naming each loop is the first step — you can't amplify what you haven't drawn on a whiteboard.
Loop Factor (k) = New Users Acquired per Cycle / Existing Users in Cycle
k
Loop factor
How many new users each existing user brings into the next cycle. k > 1 means the loop is self-sustaining; k < 1 means it decays without external input.
New Users Acquired per Cycle
Output
Net new users entering the loop in one full cycle as a result of existing-user behaviour (referrals redeemed, organic visits from indexed UGC, retargeted purchases funded by recycled margin).
Existing Users in Cycle
Input
The users whose behaviour produced the output in that cycle — typically active customers or recent purchasers.
A Shopify apparel brand runs a referral loop. In Q3, 4,200 customers received a post-purchase referral prompt. 630 referred at least one friend, and those friends placed 504 orders within 30 days.
New Users Acquired per Cycle: 504
Existing Users in Cycle: 4200
→ k = 0.12
k = 0.12 means the referral loop alone is not self-sustaining — it amplifies paid acquisition by roughly 12% per cycle but won't carry growth on its own. The lever isn't 'launch more loops'; it's getting this one to k > 0.3 by improving the prompt timing or the friend-side incentive.
Loop factor is the number to put on the wall, but cycle time is the silent multiplier. A referral loop with k = 0.2 and a 14-day cycle compounds faster than a content loop with k = 0.4 and a 9-month indexing lag. When you compare loops, plot both.
Typical loop factor (k) and cycle time by loop type in DTC e-commerce
| Loop type | Typical k | Cycle time | Primary lever |
|---|---|---|---|
| Paid → margin → paid | 0.4 – 0.8 | 30 – 60 days | Contribution margin per order |
| Post-purchase referral | 0.05 – 0.25 | 14 – 30 days | Friend-side incentive + prompt timing |
| UGC / review → social proof | 0.10 – 0.30 | 7 – 21 days | Review request automation, hero placement |
| SEO content loop (PDP + blog) | 0.20 – 0.60 | 3 – 9 months | Indexable customer-question pages |
| Email / SMS list → repeat | 0.30 – 0.70 | 30 – 90 days | Post-purchase flow, segment depth |
| Influencer / affiliate | 0.10 – 0.40 | 30 – 45 days | Creator payout structure, asset reuse |
Loops aren't a replacement for experimentation — they're what experimentation should serve. Sitting growth loops inside a broader experimentation strategy means each A/B test ladders up to amplifying a named loop: shortening its cycle, raising its k, or feeding its input. Tests without a loop usually optimise local conversion at the expense of compounding.
Growth Loops FAQ
A funnel is linear: traffic in at the top, customers out at the bottom, and you pay to refill the top every cycle. A loop is circular: the output of one cycle (a referral, a review, recycled margin) becomes the input of the next. Funnels describe a single cohort's path; loops describe how cohorts produce more cohorts.
Loops are the strategic layer; experiments are the tactical layer. Your experimentation strategy should name 1-3 priority loops, and each test should target either a higher loop factor (k), a shorter cycle time, or a larger input volume. Tests that don't map to a loop tend to optimise isolated funnel steps without compounding.
Loop factor (k) is new users acquired per cycle divided by existing users in the cycle. k > 1 means the loop is self-sustaining and growth compounds without paid input — extremely rare for DTC. Realistic targets are k = 0.15 to 0.40 for referral and UGC loops, and 0.40 to 0.80 for paid-margin loops.
Yes — in fact loops are more important for stores under €5M because paid acquisition costs eat margin faster at smaller scale. Start with a post-purchase review loop (cheapest, fastest cycle) and an email repeat-purchase loop. Referral and content loops compound later, once you have volume to seed them.
Short-cycle loops (referral, review, email) show measurable contribution in 30-60 days. SEO content loops take 3-9 months because of indexing and ranking lag. Paid-margin loops compound monthly but only if contribution margin per order is healthy — otherwise you're just laundering losses.
Run 2-3, not 6. Each loop needs instrumentation, a named owner, and dedicated test bandwidth. Stores that 'run every loop' usually run none of them well. Pick the loop with the highest k × (1 / cycle time) product, get it to a stable baseline, then add the next.
Each loop has one input metric, one output metric, and a cycle window. For a referral loop: prompts sent (input), referred orders within 30 days (output), 30-day window. You don't need multi-touch attribution to compute loop factor — you need consistent cohort definitions and a clock.
Three things, in order. Margin compression (the paid loop stalls when contribution margin drops below ad cost). Cycle-time blowout (incentives that take too long to redeem). And input starvation (the prompt that should trigger the loop is buried, mistimed, or off by default). Audit those before adding new tactics.
A viral coefficient is a specific case of loop factor — usually the k of a pure referral or share loop in SaaS. Growth loops are broader: they include paid, content, UGC, and lifecycle loops, where the 'viral' framing doesn't apply but the compounding math does.
Score each candidate loop on three axes: current k (how much it already produces), cycle time (how fast it compounds), and effort to instrument (how quickly you can measure it). The first loop to amplify is usually the one with the shortest cycle and the cleanest instrumentation — speed of learning beats theoretical ceiling.
Test ideas before you ship them
Run unlimited A/B tests, attach hypotheses to outcomes, and build a searchable archive of what works — and what doesn't.