Flywheel
A flywheel is a loop of business steps where each one feeds the next, so the same effort brings more growth every turn. The idea is Jim Collins'; Amazon made it famous.
A flywheel is a loop of business steps in which each step feeds the next, so the same effort produces a little more growth on every turn. The one-line rule: if the output of your last step doesn't make your first step easier, you have a list of tactics, not a flywheel.
The idea comes from Jim Collins, who used the image of a heavy wheel that is hard to start and hard to stop. Amazon made it famous by drawing its own version in 2001. For a small SaaS company, a flywheel is a way to answer "how does this business grow?" with a picture that shows which steps feed which, and where the friction is. This guide covers Collins' original concept, what Amazon actually drew, how a flywheel differs from a marketing funnel and a growth loop, how to draw and measure your own, examples that work at small scale, and what founders and investors are saying about flywheels in 2026.
Where the idea comes from: Jim Collins
Collins introduced the flywheel in Good to Great (2001). He asks you to picture a massive metal disk, about 30 feet across and 5,000 pounds, mounted on an axle. The first pushes barely move it. After two or three hours of persistent effort it completes one turn. You keep pushing in the same direction and it goes a little faster each time, until the wheel's own momentum starts working for you.
Each turn of the flywheel builds upon work done earlier
Jim Collins / The Flywheel Effect, excerpt from Good to Great (2001)
His point was about how breakthroughs happen. The good-to-great companies in his research could not name one defining action, one big program or one lucky break. From the outside their results looked sudden. From the inside they felt like steady accumulation in a consistent direction.
Collins also named the opposite. Comparison companies launched new programs with fanfare, didn't get quick results, changed direction, and started again. After years of lurching back and forth they never built momentum. He called this pattern the doom loop. Many SaaS founders know it as switching growth channels every quarter.
In 2019 Collins published a short follow-up, Turning the Flywheel, written after years of challenging leaders to name their specific flywheel. His challenge to readers is to name the components of their own flywheel and the order in which they drive each other: "How does our flywheel turn?"
The Amazon flywheel (and the napkin)
The best-known business flywheel is Amazon's. Collins says his monograph opens with Amazon "coming out of the dot-com bust in 2001", taking the idea he had taught them and capturing the components of its own flywheel. The fullest published account is in Brad Stone's The Everything Store (2013). Stone writes that Amazon invited Collins to a two-day management and board offsite. Drawing on Collins' concept, Bezos and his lieutenants sketched their own virtuous cycle:
- Lower prices led to more customer visits.
- More customers increased sales volume and attracted more commission-paying third-party sellers.
- That let Amazon get more out of fixed costs like fulfillment centers and the servers running the website.
- That efficiency paid for lower prices, and the loop started again.
The team's reasoning, in Stone's words, was that feeding any part of the flywheel should accelerate the whole loop. According to Stone, executives felt that after five years they finally understood their own business. When Warren Jenson asked whether to put the flywheel in his presentations to analysts, Bezos asked him not to. He considered it "the secret sauce."
A note on the napkin: many blog posts say Bezos drew the flywheel on a napkin. We could not find that detail in Stone's account or in Collins' description, which both describe a sketch made with his team at the offsite. Treat the napkin as a popular retelling. The loop itself is well documented.
What makes Amazon's version a real flywheel is that every arrow is a cause. Lower prices plausibly cause more visits. More volume plausibly lowers the cost per order. You could test each link, and the last step feeds the first.
Flywheel vs funnel vs growth loop
These three ideas overlap, and people use them loosely. The difference is what happens to the output.
- A funnel runs in one direction. Visitors become sign-ups, sign-ups become customers, and revenue comes out the bottom. Brian Balfour, Casey Winters, Kevin Kwok and Andrew Chen put the problem plainly in Growth Loops are the New Funnels (Reforge, July 31, 2018): funnels have no concept of reinvesting what comes out of the bottom, so to get more out you must keep putting more in. Funnels are still useful for measuring one step. They just don't explain how a business compounds.
- A growth loop is one specific, measurable cycle in which the output of one step becomes the input of the next, such as users creating content that brings in more users. Reforge's advice is that the fastest-growing products are usually powered by one or two major loops, and that drawing lots of small loops usually means you have a lot of weak ones. Loops are what you instrument and optimize. A viral loop is one type.
- A flywheel is the company-level picture: how product, pricing, channels, costs and customers reinforce each other. It often contains one or two growth loops. HubSpot, which publicly replaced its funnel with a flywheel, describes three things that set a flywheel's momentum: how fast it spins, how much friction it meets, and how big it is. Its marketing team's summary is that friction kills flywheels.
A practical way to hold all three: use the flywheel to explain why the business gets stronger with each turn, the loops inside it to set goals and run experiments, and funnels to diagnose a single step.
How to draw your own SaaS flywheel
Brendan Fortune, director of product management at Customer.io and an early product hire at Drip, walked Rob Walling through a version of this exercise on Startups for the Rest of Us, episode 756 (February 11, 2025, from 15:33). His version starts from customer behavior, which suits small SaaS companies well. Combined with Collins' questions, it looks like this:
- Start from your best customers, not your marketing plan. Fortune suggests picking one of your highest-paying customers and looking at what they do on a recurring basis.
- Write the recurring pattern as three to five steps. Fortune's rule is fewer than five, ideally three, even if it means cutting some corners. Each step should be a behavior or an outcome you can observe, such as "publishes a template" or "invites a teammate", not an activity of yours like "we post on LinkedIn".
- Check that every arrow is a cause. Read the loop aloud with "which causes" between steps. If a link only works because you keep pushing it by hand, it is an input, not part of the wheel.
- Close the loop. The last step must make the first step easier or bigger: more customers, more content, more data, more partners or lower costs.
- Find a price that rides the loop. Fortune's second question is whether you could charge based on that behavior. If customers pay more as they go around the loop, expansion revenue grows with the flywheel. At Drip, the loop runs from customer data to automated messages to changed user behavior and back to more data. Like many marketing automation tools, it priced on contacts, so customers who went around the loop more often paid more. See value metric and usage-based pricing.
- Put a number on every arrow. A step without a conversion rate is a guess. You will need these for the worked example below.
- Mark the weakest step and work there. Fortune's product strategy follows directly: ask why people aren't moving between steps, then prioritize making that step easier.
make sure that each step naturally leads to the next
Brendan Fortune / Startups for the Rest of Us, episode 756 (February 11, 2025)
Flywheel examples for small SaaS
You don't need Amazon's scale. Most bootstrapped SaaS flywheels are built from one of a handful of patterns.
Content
Customers or your own team create useful content, the content ranks or gets shared, readers become customers, and customers give you the questions and data for the next piece. Christopher Gimmer, co-founder of Snappa, talked through how SEO, content and YouTube can form a flywheel for a small company on MicroConf On Air (July 15, 2020). Plain content marketing becomes a flywheel only when something in the product feeds it, for example usage data that gives you original numbers to publish, or customer stories that make the next article better than a competitor's.
Templates and shared artifacts
A customer makes something in your product, shares or publishes it, and the people who see it sign up to make their own. Andrew Chen calls this the content creation loop and points to Figma and Notion as products built on it (October 6, 2026). For a small SaaS, a public template gallery is the usual version: each template is a page that can rank in search and a ready-made reason to sign up. It overlaps with user-generated content and programmatic SEO.
Integrations, partners and marketplaces
Every integration makes the product more useful to customers of the other tool and gives you a listing in its marketplace. More customers make you a more attractive integration partner, which brings more integrations. Partners work the same way. On episode 855 of Startups for the Rest of Us (October 20, 2026), Rob Walling gives Shopify as an example: thousands of agencies build their businesses on it, so a merchant who hires an e-commerce agency is likely to end up on Shopify. His test question is who has built their business or career around recommending you.
Community
Members help each other, the help makes the product more valuable, and that draws more members. Pieter Levels lists this among the moats left in 2026: users make it more valuable for each other (October 12, 2026). Community is a flywheel when member activity itself creates value, such as answers, reviews or shared data. A forum nobody visits is a cost, not a wheel. When the effect is strong enough it becomes a network effect.
Usage that expands accounts
The product spreads from one person to a team and from one team to the next, and revenue grows inside accounts you already have. Tomasz Tunguz described this as the Flywheel SaaS Company (April 19, 2015), based on conversations with executives at Asana and Atlassian: marketing brings in small customers and teams, and the sales team only upsells accounts that are already using the product. Fortune's pricing version is the same idea at a smaller scale. When pricing follows the loop, enough accounts expand to offset the ones that churn. See land and expand and expansion revenue.
Data and learning
Every customer's usage improves the product for every other customer. Walling's example on episode 855 is Stripe Radar: fraud patterns seen at one merchant protect all of them, which no single merchant could build alone. He sets three tests: it helps every customer, customers can take their data but not what you learned from it, and it keeps learning.
Every customer makes it better for every other customer.
Rob Walling / Startups for the Rest of Us, episode 855 (October 20, 2026), in the segment from 38:03
How to tell if your flywheel is actually spinning
A flywheel diagram is a hypothesis. These measures tell you whether it is true. They are our suggestions, built from the loop logic in the Reforge and Collins material above.
- Loop ratio. For a cohort of customers, how many new customers (or users, or templates) does that cohort produce by the end of one turn? A ratio of 0.3 means 100 customers bring in 30 more. Above 1 the loop grows on its own, which is rare outside strongly viral products. Most healthy SaaS flywheels sit well below 1 and still pay off, as the worked example shows. Compare the viral coefficient.
- Step conversion rates. One rate per arrow. The weakest one is where to work next.
- Cycle time. How long one turn takes. A loop with a lower ratio that turns monthly can beat a higher ratio that turns yearly.
- Share of new customers from the flywheel. Tag sign-ups that came from the loop (a template page, an integration listing, a referral) and watch whether that share rises over quarters.
- Blended CAC trend. If the wheel works, blended CAC should fall even while paid acquisition cost stays flat, because each paid customer brings others.
- Retention and expansion. Every flywheel runs through customers who stay. Watch your retention curve and NRR. If retention falls, the wheel loses mass no matter how fast the top spins.
Pick one of these as the number the whole team watches. A good North Star metric often sits on the flywheel itself, such as templates published per week or active integrations.
Worked example: a template flywheel
This is a fictional example. A small form-builder SaaS pays for some of its customers and hopes public templates will bring in the rest. It starts from 100 new paying customers acquired through ads at $300 each.
Turn one, before any changes:
- 20% of new customers publish a public template within 90 days: 100 × 20% = 20 templates.
- Each template brings 150 visitors from search in its first year: 20 × 150 = 3,000 visitors.
- 4% of visitors sign up: 3,000 × 4% = 120 sign-ups.
- 25% of sign-ups become paying customers: 120 × 25% = 30 new customers.
So the loop ratio is 30 ÷ 100 = 0.3. Those 30 customers publish templates too, producing 9 more, then 2.7, and so on. The total approaches 100 ÷ (1 − 0.3) ≈ 143 customers. The $30,000 of ad spend (100 × $300) now buys about 143 customers, a blended CAC of $30,000 ÷ 143 ≈ $210.
Fix the weakest step. Publishing is the weakest step: most customers never share. The team adds a "publish as template" button on the success screen and a short onboarding email showing other customers' templates. Publishing rises from 20% to 35%:
- 100 × 35% = 35 templates → 35 × 150 = 5,250 visitors → 5,250 × 4% = 210 sign-ups → 210 × 25% = 52.5 new customers.
- Loop ratio 0.525. Total approaches 100 ÷ (1 − 0.525) ≈ 211 customers.
- Blended CAC: $30,000 ÷ 211 ≈ $142.
The same $30,000 buys about 47% more customers (211 versus 143), and the team didn't touch the ad budget. Two caveats keep this honest. First, time: if a turn takes about six months (90 days to publish, then months to rank), most of the gain arrives over two years, not in the next quarter. Second, every rate here is an input you should measure. If templates bring 50 visitors a year instead of 150, the ratio before the fix is 0.1 and the flywheel is a small bonus, not a growth engine.
What changed in 2026
Several recent pieces from trusted voices sharpen what a flywheel is and what can break it.
- Flywheels are what moats are made of. In How to tell if a startup has a moat (October 13, 2026), Lenny Rachitsky and Dan Hockenmaier argue that only four compounding advantages matter (switching costs, network effects, scale economies and data effects), and describe each as a self-reinforcing flywheel. Speed, distribution, brand and funding are fuel that gets the car moving; a compounding advantage is the car rolling downhill on its own. Their test: look at what the team works on every day, not what it calls its strategy, because a flywheel can't be bolted on. See moat and switching costs.
- Learning loops are the strongest defense left. Episode 855 (October 20, 2026) closes on why learning loops, where every customer makes the product smarter, are the strongest of the moats that still work now that AI makes code cheap to copy (from 54:46). Einar Vollset calls it the one moat he is confident will stay ahead in the years to come.
- Agents can turn loops too. Andrew Chen argues (October 6, 2026) that AI agents can start viral and content-creation loops on a user's behalf, for example by building a shareable page and inviting people to view it. He also expects limits: people don't want their agents promoting themselves to friends, and channels may start filtering agent messages.
- Retention is under pressure. In The coming product retention crisis (September 8, 2026), Chen predicts that average product retention will collapse as apps become as quick to make as videos. For a flywheel that runs through customers who stay, that is a warning: protect the core, or the wheel loses mass.
Defensibility emerges as a byproduct of turning the flywheel over and over.
Lenny Rachitsky and Dan Hockenmaier / How to tell if a startup has a moat (October 13, 2026)
Common mistakes
- Drawing a circle of activities. "Write content → get traffic → get customers → write more content" only works if customers make the next piece of content easier or better. If the loop needs your effort at every step, it's a to-do list.
- Too many wheels. Reforge's warning applies: lots of low-powered loops usually lose to one or two strong ones. Pick one flywheel and one or two loops inside it.
- No numbers on the arrows. Without step rates you can't find the weakest step or tell whether a change worked.
- Pushing on the strongest step. More ad spend on top of a loop that leaks at "publish" buys customers who never feed the wheel. Fix the weakest link first.
- The doom loop. Collins' original warning. Changing direction every few months, before a full turn has completed, guarantees the wheel never builds momentum. Judge a flywheel over several cycle times, not one quarter.
- Pricing that fights the wheel. If customers pay the same no matter how far around the loop they go, you get the growth but not the expansion revenue. Fortune's point is that pricing and the flywheel should be linked.
- Ignoring retention. A wheel full of customers who leave after two months doesn't build momentum. Check your cohorts before celebrating top-of-loop growth.
Listen, watch and read next
- Listen: Startups for the Rest of Us, episode 756 (February 11, 2025), from 15:33. Brendan Fortune's step-by-step flywheel exercise and how to tie it to pricing, with Drip and Customer.io as examples.
- Watch: Finding Your SaaS Flywheel, Rob Walling at MicroConf 2012 (uploaded March 17, 2020). An hour on specific ways to raise revenue and cut churn in a small SaaS business.
- Watch: Using Organic Content As Your Flywheel, Christopher Gimmer on MicroConf On Air (July 15, 2020). How SEO, content and YouTube fit together into a content flywheel for a bootstrapped product.
- Listen: Startups for the Rest of Us, episode 855 (October 20, 2026), moat 7 from 38:03. The learning loop as a data flywheel, with Stripe Radar as the example.
- Read: How to tell if a startup has a moat, Lenny Rachitsky and Dan Hockenmaier (October 13, 2026). The clearest recent explanation of how flywheels turn into durable advantages.
- Book: Turning the Flywheel, Jim Collins (2019). A short monograph that starts with Amazon and walks through how to name and sequence your own flywheel's components.
Related terms
- Growth loop
- Marketing funnel
- Viral loop
- Network effects
- Moat
- Content marketing
- North Star metric
- Viral coefficient (K-factor)
- PLG (Product-led growth)
- Land and expand
Sources
- The Flywheel Effect, Jim Collins (excerpt from Good to Great, 2001)
- Turning the Flywheel and video transcript, Jim Collins (January 2019)
- Excerpt from The Everything Store, Brad Stone (2013), via Goodreads
- Growth Loops are the New Funnels, Brian Balfour, Casey Winters, Kevin Kwok and Andrew Chen, Reforge (July 31, 2018)
- The Flywheel, HubSpot
- The Innovator's Solution for SaaS Startups: The Flywheel SaaS Company, Tomasz Tunguz (April 19, 2015)
- Episode 756: Why Great Product Management Is Critical for Your Startup, Startups for the Rest of Us (February 11, 2025)
- Episode 855: The Eight SaaS Moats that Still Work in the Age of AI, Startups for the Rest of Us (October 20, 2026)
- How to tell if a startup has a moat, Lenny Rachitsky and Dan Hockenmaier (October 13, 2026)
- Will agents have network effects?, Andrew Chen (October 6, 2026)
- The coming product retention crisis, Andrew Chen (September 8, 2026)
- Moats left, Pieter Levels (October 12, 2026)
