The Anatomy of Growth

This is a narrative theme with a typological spine, synthesising how growth practitioners in this wiki — Sean Ellis, Casey Winters, Bangaly Kaba, Brian Balfour, Jackson Shuttleworth, Sarah Tavel, Adriel Frederick, Jason Cohen, Naomi Ionita, Julian Shapiro, Jag Duggal, Adam Fishman, Sri Batchu, and Elena Verna — converge on the claim that durable growth is a system, not a tactic, while disagreeing productively about where the engine sits. This theme begins where Finding Product-Market Fit ends: once you have genuine user love, the question is how to engineer compounding growth around it.

The core claim

Growth that compounds is built on self-reinforcing loops; growth that degrades is built on one-off funnels. The practitioners in this wiki reach that conclusion from different directions, but they share a foundational prior: acquisition layered on top of a product users do not love destroys value. Sean Ellis calls it the growth sequence — activation, then engagement, then referral, then revenue, then acquisition — and most companies invert it. Casey Winters puts the same point as a structural condition: ‘kindle strategies’ (non-scalable hacks) exist only to surface the ‘fire strategies’ — the viral, content, paid, or sales loops — that scale. Light kindle before you have fire and you burn resources for nothing.

The productive disagreement among practitioners is about where the compounding engine sits: in acquisition channels (growth loops, virality, product-led acquisition), in the engagement and retention layer (the leaky-bucket critique), or in the activation and onboarding ‘first mile’. These are not competing answers so much as sequencing arguments — each camp tends to locate the constraint where their own experience taught them companies fail first.

Retention is the floor, not a feature

Sarah Tavel‘s Hierarchy of Engagement is the clearest statement of the retention-first position. Three levels must be climbed in order: L1 — get users to complete the core action (the single behaviour most predictive of next-week retention); L2 — create accruing benefits and mounting loss (the product improves the more you use it; leaving costs more); L3 — self-perpetuating network effects. Most products fail at L1, not L2 or L3, and the failure is invisible: MAU growth can mask cohort decay until the bucket is already leaking hard.

Jackson Shuttleworth’s Duolingo account makes L1 operational. Four years and 600+ experiments confirmed that the streak mechanic works only atop a product people already want to use; layered onto an unloved product, a streak produces ‘Farmville’ engagement that collapses. Streak Mechanics are a retention amplifier, not a retention foundation. The load-bearing mechanism is loss aversion — once a user has built something they do not want to lose, returning becomes the default — but that aversion only attaches to things the user already cares about.

Jason Cohen’s stalled-growth diagnostic reaches the same conclusion from the SaaS churn direction. The maximum number of customers a company can ever have equals new customers added per period divided by the monthly cancellation rate. At 5% monthly churn and 100 new customers per month, the ceiling is 2,000 customers — and the ceiling is approached asymptotically, so growth stalls well before it. Logo churn is a harder constraint than most companies understand, because unlike marketing spend, cancellations scale automatically with company size. See Elephant Curve for the parallel argument about channel saturation.

The leaky-bucket critique and where acquisition fits

Naomi Ionita‘s Modern Growth Stack organises growth levers from acquisition at the top to referral at the bottom, and locates the most neglected layer in the middle: monetisation. Her claim — a 1% improvement in monetisation has 4× the bottom-line impact of a 1% improvement in acquisition — follows directly from the arithmetic: monetisation operates on the full existing user base; acquisition operates only on the marginal new user. Most growth teams over-invest in acquisition because it is measurable, attributable, and culturally exciting. The correction is not to abandon acquisition but to fill the middle of the stack first.

Jag Duggal at Nubank hardened the leaky-bucket argument into a gate: products do not scale until they pass his 50% threshold on the Bullseye Cohort variant of the Sean Ellis test. He describes the counter-intuition explicitly — the teams tempted to scale early, when overall scores are borderline, are precisely the ones who will destroy unit economics by acquiring users into a product that does not yet hold them. The Bullseye Cohort method makes the alternative concrete: find the sub-segment that already scores 70%, understand what they share, and expand the base to look like them before scaling acquisition.

Acquisition as loop, not funnel

The practitioners most focused on acquisition do not defend the funnel model — they replace it with loops. Casey Winters’s growth loop framing: a loop recycles its own output as input, so each completed cycle leaves the system stronger than before. Viral loops recycle users as recruiters; content loops recycle users as creators; paid loops recycle revenue as acquisition spend. The critical distinction is that loops compound where funnels degrade.

Julian Shapiro’s taxonomy of Product-Led Acquisition makes this structural. PLA is growth that emerges from natural product use rather than paid channels: settling debts (PayPal recipients must join to claim money), inviting to critical conversations (Slack, WhatsApp — the conversation exists only inside the product), billboarding (Hotmail footers, Calendly links, Apple hardware), and user-generated content (Quora answers ranking in search). In every case, each new user becomes a distribution node at zero marginal cost. The distinction from referral programmes is sharp: artificial incentives attract reward-seekers who churn; structural PLA attracts people who want what the existing user has.

Brian Balfour situates this within the platform cycle. Every major distribution platform — Facebook, Google, the App Store — follows the same arc: open a third-party ecosystem, give organic distribution freely to build the flywheel, then close. The companies that catch a new platform early (he argues ChatGPT is the next) get organic distribution before it becomes paid. Missing the window is not a real choice: if competitors join and customer expectations shift, abstention amounts to a strategic concession.

The activation gate: the first mile

Between the retention-first and acquisition-first camps sits a third argument: that both are downstream of activation. If users do not reach the moment of genuine value quickly, retention cannot build and acquisition is wasted.

Sean Ellis calls this ‘speed to value’ and frames it as the first step in the growth sequence — get the right people to the must-have experience before they bounce. At LogMeIn, 95% of sign-ups never completed a remote control session. Freezing the product roadmap for three months to fix the activation funnel improved the sign-up-to-usage rate from 5% to 50%. The same acquisition channels then scaled from $10,000 per month to $1 million per month with an 80% word-of-mouth mix.

Adam Fishman’s onboarding framing at Patreon is the B2B equivalent. Onboarding is the only product surface 100% of users must touch and the moment when they are most motivated. Routing high-potential creators to human coaches at sign-up improved second-month revenue by 25%; the team then productised the insight as ‘opinionated defaults’ — three-tier pricing, guardrails for the entry-point price — making best practice the path of least resistance. Counter-intuitively, better onboarding can decrease raw conversion while improving retention, because it screens out poor-fit users who would churn anyway. The First Mile Experience (Belsky’s framing) makes the psychological logic explicit: new users are lazy, vain, and selfish — they need to feel capable before they have earned capability.

Hila Qu identifies activation as almost always the largest drop-off in a PLG audit. GitLab’s aha moment — two users using two features within the first 14 days — was found by correlating every possible action against 90-day conversion and retention rates, then validated by experiment. The lesson generalises: the core action is not what product teams intuitively believe it is; it is what retention data reveals.

The adjacent user and the moving frontier

Bangaly Kaba‘s adjacent user theory adds a temporal dimension the other frameworks understate. In hypergrowth, the next cohort of users differs meaningfully from today’s base — different tech literacy, devices, cultural context, jobs to be done. Teams that dog-food products from the power-user state cannot see what is broken for the next adopter. At Instagram, declining cohort curves with no product change signalled the arrival of the adjacent user before anyone had labelled the problem. The ‘connections pivot’ — prioritising friend-to-friend connections over celebrity recommendations for new users — doubled retention over 18 months by rebuilding the activation funnel for the cohort that was actually arriving, not the one that had already converted.

Adriel Frederick frames the adjacent user as the marginal user: the person just on the cusp of taking the action you want, highest traffic, lowest conversion. Go further to the worst case — feature phone, slow connection, distant data centre — to see everything that is wrong with the experience at once. Facebook growth was not primarily clever tricks; it was grinding on fundamentals: make the product easy to find, easy to get into, easy to find your friends. Cannonballs (phone number sign-up, reliable SMS delivery, seeding the friend graph) required months of hard, unglamorous engineering.

Metrics and measurement as organisational infrastructure

Growth systems require organisational infrastructure to function. Sri Batchu’s North Star Translation Layer addresses the alignment problem at scale: a large growth organisation (300+ people at Instacart) works on diverse initiatives — checkout conversion, load time, onboarding, ads, retention — each requiring a local metric. Without a translation layer that converts each team’s local metric into expected impact on the company North Star (monthly active orders), resource allocation becomes political. With it, the question ‘should we add engineering to checkout or onboarding?’ becomes tractable: compare expected impact per engineer on a common unit.

Analytics Instrumentation (Crystal Widjaja’s framework from Gojek) is the data-layer prerequisite: the distinction between a measurement (a data point recording that something happened) and an insight (the same data point plus the contextual properties that explain why and for whom) determines whether a growth team produces decisions or dashboards. Events without properties cannot segment, hypothesise, or test.

The Growth Competency Model (Adam Fishman) is the team-layer counterpart: four quadrants — Growth Execution, Customer Knowledge, Growth Strategy, Communication and Influence — organised as a hiring and evaluation tool. Its point is that team gap-filling, not unicorn-hunting, is the structural task. Customer knowledge is the most neglected quadrant: teams with strong execution but weak customer knowledge run technically rigorous experiments that answer the wrong questions.

Where the camps converge

The retention-first camp, the acquisition-loop camp, and the activation-gate camp converge on three shared commitments.

First, growth applied before genuine product love destroys value. This is the direct link to Finding Product-Market Fit: the growth sequence presupposes that the product has passed the must-have threshold. Jag Duggal’s gate, Jackson Shuttleworth‘s streak-on-unloved-product warning, and Sean Ellis‘s activation-before-acquisition sequence all encode this prior. The Customer Love Sprints framework (Noah Weiss at Slack) is a recovery tool — a rapid cycle of 15–25 fixable fixes shipped in two to four weeks — for teams whose satisfaction metrics have fallen precisely because they scaled before the product held.

Second, channel saturation is inevitable. Jason Cohen’s Elephant Curve applies to every acquisition channel: audience saturation, channel decline, and competitive crowding pull the curve down after its S-curve peak. Growth that depends on a single channel is time-limited by structural forces no team can reverse. Durable growth requires either building structural PLA into the product’s architecture or continuously identifying and seeding new channels before the current ones peak.

Third, the speed of the product’s own movement determines the applicable playbook. Elena Verna at Lovable describes the AI-native case: in a category moving faster than any funnel tweak can compound, the growth lever is not improving existing journeys but shipping new loops. PMF must be recaptured every three months when what the product can do changes monthly. The toolkit described by every other practitioner in this theme still applies — it just runs at a different cadence.