Breaking the Rules of Growth: Why Shopify Bans KPIs, Optimizes for Churn, and Prioritizes Intuition
- Aisha Washington
- 1 hour ago
- 7 min read
Shopify operates at a scale that makes its unconventional growth philosophy difficult to dismiss. In this conversation, Archie Abrams, Shopify’s VP of Product and Head of Growth, describes an organization serving a substantial share of American e-commerce while resisting several habits that have become standard across software companies.
The apparent contradictions are the point. Shopify accepts high merchant churn, distrusts locally optimized conversion rates, follows experiments for years, and sometimes ships statistically neutral changes because the new experience simply seems better. Behind these choices is a coherent idea: growth should reflect durable merchant success, not whatever number moves most conveniently this quarter.
Shopify’s Business Model Changes the Growth Equation
Abrams frames Shopify as infrastructure that shoppers often use without realizing it. He notes that the platform powers roughly 10% of US e-commerce and processed about $235 billion in global gross merchandise volume in 2023. That scale resembles an economy more than a conventional software customer base.
Yet Shopify does not treat every new merchant as a subscription that must be retained indefinitely. Its mission is to expand entrepreneurship, which means reducing the cost and difficulty of trying to build a business. Many of those attempts will fail. From Shopify’s perspective, that outcome does not necessarily indicate a defective funnel.
The economics help explain why. Subscription revenue matters, but Abrams says most of Shopify’s revenue comes from payments associated with merchant sales. When a merchant succeeds, Shopify participates in that growth. A small number of exceptional businesses can therefore contribute more than a large population of modest subscribers.
Abrams compares the pattern to an investment portfolio in which a few outsized winners determine the overall return. Shopify is not literally trying to make merchants leave. It is accepting that broad access produces many unsuccessful attempts alongside a smaller number of businesses with enormous impact.
That leads to a different unit of analysis: the total GMV generated over time by merchants who joined during a particular period. Instead of asking only how many accounts remain active, the company asks how much lasting commercial activity the cohort creates.
Why Retention Can Be the Wrong Target
Most SaaS playbooks treat churn as an enemy. That is sensible when customer acquisition is expensive and revenue depends mainly on recurring fees. Shopify’s combination of brand awareness, word of mouth, and transaction-based monetization makes the calculation less straightforward.
Trying to maximize retention could encourage Shopify to admit only merchants who already look likely to survive. That would improve a retention percentage while excluding uncertain founders who might eventually build important businesses. It could also conflict with the company’s stated ambition to make entrepreneurship more accessible.
This distinction matters beyond Shopify. A metric should represent the value a company intends to create; it should not become the mission by default. If the goal is to help more people attempt entrepreneurship, then some conventional SaaS measurements may penalize the very openness the product is designed to provide.
Shopify still examines retention, revenue, gross profit, and other operational signals. Abrams’s argument is not that measurement is useless. It is that no isolated KPI can adequately express the value of a merchant cohort whose most consequential outcomes may take years to emerge.
Measure How Many People Succeed, Not Just the Conversion Rate
Abrams identifies a common problem in large growth organizations: every team receives one section of the funnel and is asked to improve its conversion rate. This arrangement appears rigorous, but it can reward behavior that damages the wider journey.
Suppose a team owns conversion from onboarding to activation. One easy way to raise that rate is to make the preceding step more demanding. Fewer people enter onboarding, but those who remain have stronger intent. The percentage improves even though fewer customers may reach activation overall.
Shopify instead emphasizes absolute outcomes: how many additional merchants completed the journey, began selling, or generated long-term cohort value? This changes the team’s incentive. Removing signup friction might lower the percentage of registrants who survive each later stage, yet still produce more successful merchants in total.
The broader economics may improve as well. A wider top of funnel can reduce customer acquisition cost even when retention rate or average lifetime value declines. For Abrams, the correct question is not whether a local ratio rose. It is whether the work created an incremental population of productive merchants.
Experiments Need a Much Longer Memory
Perhaps Shopify’s most transferable practice is its use of long-term experiment holdouts. A test may be declared after a few weeks so the team can keep shipping, but the original groups remain available for later analysis. Shopify’s experimentation systems revisit outcomes after three, six, nine, and 12 months, with some comparisons continuing for several years.
The company uses two complementary designs. A portion of users—Abrams cites a 5% quarterly holdout—is excluded from a collection of changes. For experiences limited to new merchants, teams may begin with an even split, release the winning version broadly, and continue observing the cohorts originally assigned to each condition.
This delayed measurement frequently revises the initial story. Abrams estimates that 30% to 40% of tests showing early improvement no longer produce incremental value after a year. Often, the intervention merely accelerated an action that would have happened anyway.
A payment-failure notification illustrates the problem. Reminders initially appeared to preserve more merchants, but the long-term advantage disappeared. The feature had prompted some people to update their billing details without changing their eventual commitment to building a business.
The opposite pattern is less common but strategically important. Preconfigured content blocks for online stores did not immediately increase conversion to paid plans. Six months later, however, merchants who received them were more likely to be selling and producing GMV. Better-looking, more effective storefronts may have helped them secure early sales, creating momentum that a short experiment could not detect.
For teams unable to wait a year, Abrams recommends instrumenting the furthest downstream signals available and treating early wins with restraint. A positive short-term result may justify shipping, but it should not automatically be counted as permanent business value.
Lowering Monetary Friction Can Reveal Future Winners
Trials, introductory pricing, incentives, and useful credits all fall under what Abrams calls monetary friction. These mechanisms matter because early-stage entrepreneurs often have expenses before they have revenue.
Reducing that pressure does more than shift the date of payment. In some cases, it gives a founder enough time to complete a store, test a proposition, and make an initial sale. That experience can causally alter whether the business survives.
This is another reason short-term segmentation can mislead. A merchant who looks commercially weak during onboarding may later become one of the valuable outliers in a cohort. Designing only for people who can pay immediately risks selecting for present resources rather than future business potential.
Abrams also highlights personalization during signup and onboarding. Collecting relevant information allows Shopify to offer guidance suited to a merchant’s situation. The aim is not to add more steps for their own sake, but to build trust and provide enough direction for users to reach a meaningful first success.
Growth Is Organized Around the Entire Merchant Journey
Shopify’s growth organization combines two major groups: Growth R&D and Growth Marketing. Marketing covers familiar acquisition disciplines such as paid media, affiliates, email, content, and SEO. Its job is largely to bring prospective merchants to Shopify.
Growth R&D includes product, design, engineering, data, support, and internal growth infrastructure. Growth Product works across landing pages, signup, onboarding, monetization, and engagement. Another pillar develops shared systems for experimentation, communications, business intelligence, and marketing technology. Customer Support contributes both adviser tools and merchant-facing experiences such as the Help Center and AI-assisted self-service.
This structure connects acquisition with activation and ongoing support. Goals are built around forecasted organic performance, expected incremental lift, payback constraints, and the multi-year value of acquired cohorts. Individual channels still operate within LTV and CAC guardrails, but product teams are expected to demonstrate additional cohort value rather than celebrate disconnected funnel improvements.
Where Metrics End, Product Judgment Begins
Shopify’s product philosophy does not demand a statistically significant win before every release. Abrams says that when two options test as neutral, the company may choose the version it would prefer if no incumbent design existed.
That decision relies on judgment—what the conversation calls taste or intuition. It does not mean ignoring evidence. Experiments establish that the change is unlikely to cause meaningful harm; product judgment then decides which experience is clearer, more coherent, or more useful for merchants.
This approach counters the status quo advantage built into experimentation. If an existing design and a more thoughtful replacement perform similarly, insisting that the replacement prove an immediate numerical lift can freeze products in place. Shopify leaves room for teams to improve quality even when the benefit is difficult to capture within a short measurement window.
Building for a Hundred Years
Abrams connects these operating choices to CEO Tobi Lütke’s ambition to build a hundred-year company. Shopify’s core product teams consider what commerce may require over decades, while merchant services and growth work on different horizons and problems.
Long-term thinking also shapes technical decisions. Abrams describes recurring reviews in which R&D leaders examine projects with Lütke, including whether the architecture preserves future options. The underlying belief is that technical architecture constrains strategy: expedient systems can quietly narrow what a company will be able to build later.
The same logic explains Shopify’s continued focus on entrepreneurs and small businesses. Today’s dominant brands may not remain dominant for another generation. If Shopify helps new businesses start and grow, it has the opportunity to develop alongside tomorrow’s important merchants.
Shopify’s model cannot be copied mechanically. Most companies do not share its scale, monetization, or ability to maintain multi-year holdouts. The deeper lesson is more useful: select metrics that reflect the actual economics, design incentives around total outcomes, and keep checking whether apparent wins survive contact with time. Growth becomes more credible when teams are willing to learn that yesterday’s success was temporary—or that an overlooked improvement was quietly compounding all along.
