business / growth-and-scaling · article
business/ · 2,510 w · 11 min · ✎ dialogue

Growth and Scaling

Growth is not a single decision but a continuous series of choices about what to prioritize, what to ignore, and when to change course entirely. The founders who scale well tend to share a counterintuitive quality: they are comfortable letting problems go unsolved, they resist the temptation to grow before they're ready, and at some point they're willing to cannibalize their best current business to build a better future one. This article draws on stories from Zuckerberg, Kalanick, Collison, Silberman, Preston-Werner, and case studies from Fortnite, Starbucks, and The Diff to lay out what scaling actually looks like.

The Early Stage: Deliberate, Not Explosive

The canonical narrative of startup growth is a hockey stick — sudden, exponential, unstoppable. The reality for most great companies was slower and more deliberate.

Facebook grew school-by-school. It took a year to reach 1 million users, which at the time felt fast. But Zuckerberg's account of early growth emphasizes intentionality: they only launched at a new school when they had enough ad revenue to pay for the servers. There was no VC-subsidized blitzscaling — the expansion was paced by the business's ability to pay for itself. The side effect of this constraint was cleaner data, real identity, and a product that baked properly before it spread.

When asked about the quality-speed tradeoff in early Facebook, Zuckerberg put it simply: "You can't 80/20 everything. There are some things you have to go beyond that and be the best in the world at." The decision to require real school email addresses — manual and slow — was one of those things. Dustin Moskovitz wanted to drop the requirement to grow faster. Zuckerberg held firm. The debate set a tone for the company's early culture around data quality.

Stripe was even slower at the start. Patrick Collison showed a transaction volume chart from Stripe's first six months: basically zero. Their first production user was a friend named Ross. They operated in invite-only beta for over a year — what a friend called the "baby blanket" of startups. Fifteen months passed between their first user and their public launch with 10 people. This wasn't a failure; it was a deliberate choice to build something that worked reliably before exposing it to the world. The lesson Collison drew: doubt doesn't go away with early success. It's persistent, even in the most successful companies.

Pinterest at four months post-launch had just under 3,000 users. Ben Silberman's description of the startup journey: "It's not like a marathon. It's actually a roadtrip at night with no headlights. You think you're going to Toledo but you're actually going to Miami and you might not have enough gas." The pivot from "building what he wanted" to "finding people like me" — identifying the niche of passionate early adopters rather than chasing the mass market — was what ultimately unlocked growth.

GitHub launched with a deliberately simple mission: "git hosting — no longer a pain in the ass." Tom Preston-Werner's core belief: "Don't worry too much about when you enter a market. In every market, most products are terrible." Timing matters less than execution quality and the right team.

Let Fires Burn: Intelligent Triage at Scale

Once growth starts, founders face a new problem: more fires than they can fight. The instinct is to put out every fire. The discipline is knowing which ones to let burn.

Reid Hoffman, reflecting on PayPal's early days, describes the deliberate choice to ignore customer complaints as the user base scaled exponentially. The customer service team was three people managing 10,000 emails a month. Phones rang 24 hours a day. PayPal turned off the ringers on desk phones and used cell phones instead. The logic: "We have to treat the future customers, not just the current ones. If all we did was the current ones, we'd never get to the future customers." Eventually, they flew to Omaha, set up a call center, and built a 200-person customer service department in two months — solving the problem all at once rather than imperfectly throughout.

The complementary principle comes from SurveyMonkey's growth under Selina Tobaccowala: SurveyMonkey's website was widely considered ugly. Users complained. Design critics shuddered. Tobaccowala let it burn for years, choosing instead to invest in engineering, reliability, and international expansion. Her reasoning: "When you're trying to get somebody through a difficult experience in something more functional, it's far more important that the product is easy to use — and beautiful and easy to use are not always the same thing." Amazon is the canonical large-company example of this logic — the website wins no design awards, but it relentlessly optimizes for convenience, price, and speed.

VMware's Jerry Chen went further, tracking complaint volume at launch as a positive metric: "You're releasing while you're still embarrassed, and people only complain when they care."

Kuli Kuli said yes to Whole Foods' request for a nationwide launch in January — six months away, in June — despite not being ready. Their operations lead started shaking her head. Lisa Curtis said yes anyway. The launch was botched: moringa clogged factory machines, a snowstorm delayed the first shipment, the texture was gritty. But the outcome was Kuli Kuli in 3,000 stores, quadrupling its footprint in a year. The unforced errors were real; the growth was real too. "I don't think I would've changed saying yes and seizing the opportunity."

The framework: triage fires by probability of occurrence, severity of damage, and correctability. A potential Chernobyl (e.g., SurveyMonkey's lack of database backup) that has a 0.1% daily probability is different from one with a 1% daily probability. The former can wait a few months; the latter cannot.

The Growth Team: Small Details Compound

Zuckerberg's description of Facebook's Growth Team captures how unglamorous real growth work is. The team's job was to help people get their friends onto Facebook — "relatively mundane things," in Zuckerberg's words.

One example: depending on how email contacts were sorted in the invite flow, the number of invites sent could vary by 3–4x. The bug that caused certain users' contacts to show blank names — traced to a specific email client used in certain countries — was fixed by one engineer, and it measurably moved Facebook's growth. "This is not product innovation, it's marketing." But it compounds.

Uber's approach to supply-side growth shows the same attention to non-obvious detail. Rather than showing drivers a heatmap of raw demand (which would cause everyone to cluster in the same spot), Travis Kalanick's team showed residual demand — demand minus current supply. The math department's job was to keep pickup times short while keeping utilization rates high. Small optimizations to the algorithm translated directly into better driver economics and better rider experience.

Kalanick also made the point that Uber's early growth was entirely word-of-mouth. Zero marketing spend with 120 employees. The product was good enough that 50% of users who had ever ridden used Uber in the last 30 days — a retention rate that made virality cheap. "When I'm having a bad day, I just go out to our overall revenue graph," he said, showing a hockey-stick chart. The early signal that justified the growth bet wasn't global — it was seeing cities outside SF (Seattle, New York, Chicago) launch higher than SF had in their early days.

Bet-the-Company Decisions: When Good Isn't Good Enough

Every mature company eventually faces a version of the same question: if I were building this from scratch to serve the same customer need, would I build what I run today? When the honest answer is no, it's time for a bet-the-company move. These are late-stage pivots with measurable opportunity costs on the line.

Meta's mobile pivot is the gold standard case. When Facebook began prioritizing mobile, it was growing desktop revenue at 88% with 42% incremental operating margins. The desktop advertising business was among the best ever built. Shifting to mobile cannibalized that business — mobile had worse UX, no monetization, and far lower ad yields. From the outside, it looked like a company sabotaging its own golden goose. From the inside, Zuckerberg believed desktop was a local maximum. He was right. Shares returned 25% annualized from IPO to the next bet-the-company move (the Meta/Metaverse rebrand in 2021).

IBM's System/360 is the historical precedent. IBM spent $5 billion in mid-1960s dollars — equivalent to roughly $164 billion today — to standardize its entire computer product line on a single instruction set. IBM held roughly two-thirds of the global computer market at the time and was growing profitably. The System/360 required IBM to build all peripherals simultaneously (standardization without a complete ecosystem is useless), and ultimately forced IBM to become the world's largest manufacturer of integrated circuits. The project was almost abandoned; one of the classic books on software project management (The Mythical Man-Month) was essentially a case study of System/360's difficulties. It succeeded, and cemented IBM's dominance for a generation.

Netflix's streaming pivot had an 80%+ stock drawdown over six months in late 2011, when they launched and then killed a separate DVD-only brand, and created separate pricing for streaming. Revenue growth slowed from 30% to 12%; profits dropped more than 90%. But this is, as The Diff's Byrne Hobart writes, "the signature of a well-timed bet-the-company choice — new models are purely accretive only when they've been perfected somewhere else."

OpenAI made a double pivot in late 2018 and early 2019: from robotics to transformers technologically, and from nonprofit to capped-profit organizationally. The organizational shift was driven by necessity — once they identified transformers as the path, the capital requirements were incompatible with a nonprofit funding model.

A key insight from all of these: you can never truly bet the entire company. IBM used its existing customer relationships and technical expertise. Meta seeded Metaverse accounts with data from Facebook. Netflix used its subscriber base to cross-sell streaming to DVD customers. OpenAI retained the talent it had aggregated. The assets that carry over are relationships, people, brand, and data.

Fortnite: A Growth Design Case Study

Fortnite earned $2.4 billion in 2018 — the most annual revenue of any game in history at the time. The growth model was deliberately engineered around several interlocking mechanics:

Free-to-play with cosmetic microtransactions. The base game is free, which removes the barrier to entry entirely. Revenue comes from skins and other cosmetics — status signals that don't affect gameplay. This separates willingness to pay from competitive advantage, which was the specific objection to earlier pay-to-win games.

Seasonal Battle Pass with expiration. The Battle Pass is time-limited. Items expire at the end of the season. This creates FOMO urgency — players who want the exclusive content must engage consistently during the season rather than accumulating items over time.

Social and viral loops. Players recruit friends to play together. The game is better with people you know, which creates social pressure to onboard your network. The word-of-mouth loop was free.

The model is instructive beyond gaming: free entry + paid status + artificial scarcity + social proof is a growth architecture that has worked across many categories.

Recession Playbook: Starbucks 2008–2010

Howard Schultz returned to Starbucks as CEO during the 2008 recession and executed a turnaround that became a case study in how to grow through a downturn rather than merely survive it.

Listen through structured channels. Schultz built MyStarbucksIdea.com — an online suggestion platform inspired by Dell's IdeaStorm — in 63 days, over Christmas. Over 100,000 suggestions came in; 100 were implemented. The platform served both as a listening mechanism and a retention signal to customers that Starbucks was paying attention.

Preserve the core. Internal research showed customers weren't abandoning Starbucks entirely — they were coming less frequently. Core customers were ordering customized drinks and paying for extras. Costco CEO Jim Sinegal's advice to Schultz: "Protect and preserve your core customers. The cost of losing your core customers and trying to get them back during a down economy will be much greater than the cost of investing in them." The response was the Starbucks Rewards Card — free refills on brewed coffee, free customizations — which strengthened retention without cutting prices or quality. As of 2022, the program carries over $2 billion in unspent card balances.

Process innovation over cost-cutting. The board pushed for aggressive layoffs, lower coffee bean quality, and eliminating the healthcare program. Schultz refused. Instead, he overhauled operations:

  • New POS systems saved 700,000 annual hours of wait time
  • Supply chain went from 3/10 to 9/10 on-time deliveries
  • New scheduling software for store managers automated coordination of barista availability with store traffic
  • Total permanent annual savings: $580 million

Play offense and defense simultaneously. Companies that used a "progressive" strategy — balancing offensive investment and defensive cost management — had a 37% probability of leading their markets post-recession. Defense-only companies had a 21% chance. Starbucks did both: it cut costs through process innovation (defense) while launching new products including Via instant coffee, which hit $100M in U.S. sales within 10 months of launch (offense). Starbucks ended up in the 9% of businesses that flourished post-recession, posting record revenue of $10.7 billion in 2010.

On People and Culture

Tom Preston-Werner's framing at YC Startup School distills a core truth about scaling: "A company is nothing more than the decisions it makes. Decisions are made by people. The only thing that matters is people." When you hire someone, the question isn't just whether they can do the job — it's "How are they going to push the company forward?"

Zuckerberg's lesson from the Yahoo $1B offer moment was about people, not money. He felt totally misaligned with the management team he'd hired — people who had been brought in to build a company that would get sold, not one that would reshape how people share information. "You don't want to hire people just like you — but people who are aligned with you." The misalignment, not the offer itself, was the painful part. He replaced most of the management team in the following year.

His prescription for maintaining culture at scale: keep the company as small as possible relative to its footprint. "The best answer for how to not suffer from your company being big is to not let it get big." The goal is an extreme users-per-employee ratio — staying in the part of the curve where engineers are each responsible for a large number of users, rather than drifting toward the enterprise end where headcount and output decouple.

Thread · 0 replies+ add reply
no replies yet — be the first to write back to this article.