Growth loops explained simply

Why loops matter more than funnels

If you've ever stared at a funnel chart and wondered why growth flatlines the moment paid spend pauses, you've felt the answer: funnels are linear, loops compound. A funnel takes a user from top to bottom once. A growth loop reinvests the output of one step as the input of the next — and the same user (or the content, revenue, or invite they produced) keeps fueling acquisition without you topping up the budget.

Reforge, Andrew Chen, and Brian Balfour have pushed this distinction for years because it changes how you build dashboards and write OKRs. You stop asking "how do we improve conversion at step three?" and start asking "what is the shortest cycle time and highest output per input we can sustain?". For a growth analyst or PM, fluency with loops shows up in interview prompts and roadmap reviews.

The five loop types you should know

There are dozens of taxonomies floating around, but five categories cover roughly 95% of what you'll see in real companies and in interview questions. Each has a different input, output, and dominant constraint.

Most B2C consumer products lean on the first two; B2B SaaS leans on paid plus sales; marketplaces typically need at least three running in parallel to defend a moat.

A viral loop is the cleanest example. A user signs up, invites N friends, some fraction convert, and the cycle repeats. Dropbox's "give 250MB, get 250MB" referral is the canonical case study — it turned every existing user into a free acquisition channel with a measurable payback.

A content loop uses the user's own creation as the bait. A Pinterest user pins boards, those boards rank in Google, searchers click through, some sign up, and they create new pins. Medium, StackOverflow, and Reddit run variants of this loop. The constraint is whether each new user produces enough indexable, durable content to seed the next batch of searchers.

A paid loop is the most misunderstood. It's only a loop if LTV > CAC and the company actually reinvests revenue into more acquisition. Without reinvestment, you have a funnel with extra steps. Uber and DoorDash burned cash for years specifically to keep this loop spinning until the unit economics caught up.

A sales loop is the B2B classic: closed revenue funds more SDRs and AEs, who close more revenue. It requires high ACV and disciplined ramp, otherwise hiring outpaces productivity and the loop stalls.

A UGC engagement loop is what makes Instagram and TikTok feel inescapable — content from existing users keeps them coming back, which keeps them creating more. The metric is network density, not headcount.

Loop math — what makes one strong

Every loop reduces to roughly the same equation. The shorthand most growth teams use is:

growth_rate ≈ (output_per_input × conversion_rate) / cycle_time

Three levers, every one of them measurable. Output per input is how many new things (users, dollars, pieces of content) come out the other side. Conversion rate is the survival probability of each step inside the loop. Cycle time is how long one rotation takes — and it's the lever most teams underinvest in.

For the viral case specifically, the canonical formula is the K-factor:

K = invitations_sent_per_user × invite_conversion_rate

• K > 1: each user yields more than one new user → exponential growth
• K = 1: steady state, just replacing churn
• K < 1: the loop alone cannot sustain growth and needs another channel

A nasty trap here is that K measured on week one is almost always inflated. Early users are enthusiasts; their invites convert at rates a normal cohort never reaches. Reforge teams recommend tracking K on rolling 30-day cohorts and watching for the inevitable decay.

The same logic applies to non-viral loops. A content loop with output of 3 SEO-indexable posts per new user, a 2% search-to-signup conversion, and a 90-day cycle time behaves very differently from a loop with one post per user and a 365-day cycle, even if the conversion rates match.

Real-world loop teardowns

Looking at how mature companies stack loops is the fastest way to internalize the model. None of these companies run a single loop — they run portfolios.

Notion runs at least three loops in parallel. The first is a content loop: users publish templates, those templates rank for long-tail searches like "notion CRM template", searchers sign up, some become creators. The second is a viral loop through shared workspaces — invite a teammate to view a doc and they're one click from creating their own. The third is a paid loop funded by enterprise expansion revenue, reinvested into outbound and partnerships.

Airbnb rode a paid loop hard while SEO content quietly compounded in the background. The infamous Craigslist double-post trick was a content loop hack: every host listing produced a cross-post that pulled traffic back. Once organic SEO kicked in, paid became a discretionary accelerator.

Tinder runs a viral loop (invite friends → more matches → engagement → invites again) on top of a paid loop funded by premium tiers. The interesting bit is cycle time: a match in seconds, an invite in hours. That's why Tinder outgrew dating sites with longer cycles.

Common pitfalls

When teams try to operationalize loops for the first time, the most common failure is treating paid acquisition as if it were a loop on its own. Pumping money into ads is not a loop unless the revenue from acquired users meaningfully funds the next round of spend. Without that reinvestment connection — and without LTV exceeding fully-loaded CAC — you have a funnel that bleeds cash whenever the marketing director takes vacation. The fix is to model the reinvestment explicitly and stop calling it a loop until LTV/CAC clears a defensible threshold like 3:1 at 18-month payback.

A second pitfall is ignoring cycle time in favor of conversion rate. Teams optimize the invite landing page from 8% to 11% conversion and celebrate, while the actual cycle from signup to invite-sent takes 14 days because the prompt is buried in a settings menu. Halving the cycle time gives you the same compounding benefit as doubling conversion, and it's often easier — surface the loop earlier, prompt at the right moment, remove friction between rotations.

A third trap is measuring loops with funnel dashboards. A funnel dashboard answers "how many users got from step A to step B last week?". A loop dashboard has to answer "of the users who entered the loop in March, how many had completed at least one full rotation by May, and what's the average output per completed rotation?". These are different SQL queries, different visualizations, and often different tables entirely. If your growth team only has funnel dashboards, you can't actually tell whether your loop is healthy.

A fourth and surprisingly common pitfall is mistaking a one-off campaign for a loop. A referral promotion that gives a $20 credit for a successful invite will spike acquisition for two weeks and then collapse — that's a campaign, not a loop, because the input doesn't replenish from the output. Related: running only one loop and calling it a strategy is fragile. If Google changes its algorithm and your content loop dies, you have nothing to fall back on. Resilient growth models stack at least two independent loops.

Interview answers

These questions show up in growth PM, growth analyst, and senior product analytics interviews. Crisp answers matter more than length.

"What is a growth loop?" A self-reinforcing cycle where the output of one step becomes the input of the next, producing compounding growth without continuous external input. Contrast with funnels, which take a user through linearly and end.

"Loop versus funnel — what's the difference?" Funnels are linear and single-pass; loops are circular and compound. Funnels improve via per-step conversion lifts; loops improve via output per input, conversion, and cycle time. You can model a loop as a series of funnels, but the defining feature is the feedback connection back to input.

"Give me three loop examples." Dropbox referral (viral, K-factor driven), Pinterest pins ranking in Google (content), Airbnb reinvesting host revenue into paid acquisition (paid).

"How would you measure a loop?" Track inputs, conversion at each internal step, cycle time per rotation, and output per input. Build a dedicated loop dashboard separate from funnel dashboards, with cohort views so you can spot decay.

"How would you improve a slow loop?" Cut cycle time before lifting conversion — usually easier and the impact compounds the same way. Surface the loop trigger earlier, reduce friction between rotations, and instrument the moment a user could rotate but doesn't.

If you want to drill questions like these, NAILDD has growth-focused PM and analytics sets covering loops, funnels, and the math underneath.

Related reading

• Growth loops primer for PMs
• AARRR pirate metrics framework
• How to calculate K-factor in SQL
• North Star metric for PM
• Growth PM vs regular product manager

FAQ

Is one loop enough for a growing startup?

Early on, yes — most companies start with a single dominant loop while everything else is noise. Dropbox lived on its viral loop for years before paid acquisition mattered. But as you scale past a few million users or into new segments, the original loop typically decays or saturates. By that point you want at least two independent loops so a shock to one doesn't kill growth, and so you can target different audiences with different mechanics. Most $1B+ businesses have three or more.

Do growth loops work for B2B?

They do, but the mechanics look different. Viral loops are weak in B2B because invite networks are tighter and longer-cycle. The strong B2B loops are content (technical SEO, developer docs, case studies), sales (revenue funds more reps), and product-led recommendations where users carry the product across companies as they switch jobs. Stripe, Datadog, and Notion all combine these. The cycle times are longer — sometimes years — but the conversion rates are correspondingly higher.

Can a company grow without paid acquisition at all?

Yes. Pinterest, Wikipedia, and early Reddit all grew with effectively zero paid spend by leaning entirely on content and UGC loops. It's harder and slower, and it requires a product where each user produces durable, indexable output. If your product doesn't generate content or natural invitations, you'll need paid as either a primary loop or a topping-up channel. The decision usually rests on whether your unit economics support a paid loop with LTV at least 3x fully-loaded CAC.

How is K-factor different from a growth loop?

K-factor is a specific metric inside a viral loop — it measures how many new users each existing user generates through invitations. A growth loop is the broader concept that includes viral, content, paid, sales, and UGC variants. Every viral loop has a K-factor; not every growth loop does. Think of K-factor as the "tightness" parameter of one specific loop type.

What's the fastest signal that a loop is dying?

Cycle time stretching out and output per input drifting down, in that order. A healthy loop has a stable or shrinking cycle time and a roughly constant output per cohort. When you see new cohorts taking longer to complete a rotation or producing fewer invites/posts/dollars per user, you're looking at decay — usually from saturation, channel changes, or shifting user mix. Catching it early gives you months to launch a second loop before growth stalls.

Where do most teams go wrong when first modeling a loop?

They draw a beautiful diagram and never instrument it. The hard part is the SQL, dashboards, and cohort views that show whether the loop is rotating, how fast, and with what output. If you can't pull "cohort entered loop in March, completed rotation by May, produced X outputs" in one query, you don't have a loop — you have a hypothesis.