v1.2
THE FORMULA Updated 18 Nov 2026 Updated weekly

The Formula.

What the big bets and the breakout brands have in common, and how to use it before your next move.

What this is

300 Big bets in the study, each one costed and sourced
89 Founders, leaders and companies who made them
224 Bets checked against sources with a result we can judge. 104 are fully checked

This is an ongoing study of the big bets founders and leaders make, and how brands break out. Next to the bets we study brands year by year, the ones that broke out and the ones that rose and fell, straight from the numbers. Every number on this page comes from that study, and every example links to its source. It grows every week, and this page changes with it.

Checked means at least one reputable report backs it. Fully checked means a company filing, an annual report or an official release, or two reputable reports that agree. Anything we couldn't check, or found disputed, is left out.

In one line: back the bet people doubt when you can say what they're missing, line up the money and the patience first, keep every promise to your customer, and steer clear of the red flags.

Does it hold up? Across the checked bets, the ones people doubted won 81% of the time, against 31% for bets most people backed. Bets with none of the red flags won 77% of the time, against 27% for bets with one. So far we've looked at bets after we knew how they ended. Next we test it properly: we score 10 new brands using the formula, write down our call, and only then check what happened.

The rules

  1. 01

    Back what people doubt, when you can say what they're missing.

    Bets most people doubted at the time won 81% of the time (67 checked bets). Bets most people backed won 31% (58). On the fully checked bets it is 14 of 18 against 8 of 29.

    ExampleTony Fadell's startup Fuse was turned down by investors. In 2001 he took an Apple contract and built the iPod. It sold $143m in its first full year and $9,153m in 2008.

    Apple 10-K 2002 ↗ · Apple 10-K 2008 ↗ · CNBC ↗

  2. 02

    Make a new category if you can.

    Bets that made a new category won 82% of the time and won big 68% of the time (40 bets). Across every checked bet, 36% won big.

    ExampleSpotify launched free streaming in 2008 to win people back from piracy. It was still losing money in 2017, €1,235m that year, and listed in April 2018 at about $26.5bn.

    Spotify F-1 ↗ · Al Jazeera ↗

  3. 03

    Say no, and know what it makes room for.

    Bets built on saying no, to a buyer, to easy volume or to the discount, won 83% of the time (30 bets). Only one has been lost outright. On the fully checked bets it is 11 of 11.

    ExampleHermès said no to LVMH and locked 50.2% of its shares into a family holding in 2011. Sales went from €2,400.8m in 2010 to €16,002m in 2025, with a 41% margin.

    FashionNetwork / Reuters ↗ · Hermès 2010 results ↗ · Hermès 2025 results ↗

  4. 04

    Never break a promise to your customer.

    All 3 checked bets that broke a promise customers relied on have lost. Breaking trust is the main reason behind 12 of the 89 bets that lost or came out mixed.

    ExampleIn 2011 Netflix raised prices and tried to split its DVD service off as Qwikster. It lost 805,000 US subscribers in a quarter and its shares fell 35% in a day.

    Netflix 10-Q ↗ · CNN Money ↗

  5. 05

    Have the money and the patience lined up before you start.

    Bets made with patient money and a long horizon won 76% of the time, and only 1 in 70 was lost outright. Winning bets took a median 22 months to pay off (97 bets). Big wins took 28 months (53).

    ExampleIn 1994 Patagonia decided to move all its cotton to organic. It cut its sportswear line from 166 products to 66, and says it took two years for sportswear sales to come back.

    Patagonia ↗

  6. 06

    Watch the red flags.

    Scaling before you have proof, drifting off your core, leaning on money that can leave, one key person with no plan B, taking the same playbook to a new country. Bets with any red flag won 27% of the time (56 bets). Bets with none won 77% (168). Scaling before proof is the worst: 2 wins in 15. The same playbook in a new country: 2 wins in 7.

    ExampleIn 2014 Nest recalled about 440,000 Protect smoke alarms because a wave-to-silence feature could be set off by accident and delay a real alarm. No one was hurt.

    US CPSC ↗

  7. 07

    Expect the breakout to take years, and expect it to be quiet.

    Winners took a median 7.5 years to break out. With the founder in charge it was about 6 years. Without, about 32. 17 of 50 brands with a long enough sales record doubled in three to five years without a single year of 40% growth.

    ExampleMoncler more than doubled its sales from €580.6m in 2013 to €1,194m in 2017, and never grew more than 27% in any one year.

    Moncler 2013 results ↗ · Moncler Annual Report 2025 ↗

  8. 08

    After the breakout, keep telling the story, grow in steps and raise the price.

    24 of 53 winners kept telling their story after the breakout, against 1 of 15 brands that rose and fell. 38 of 53 winners grew in steps, against 6 of 15. Winners raised the price of their hero product after the breakout. Brands that stalled held it still or cut it. Sweetgreen raised menu prices from 2021. In 2025 sales at its existing stores fell 7.9%, and in December it brought in a $10 value bowl.

    ExampleArc'teryx's Alpha SV jacket went from $749 in 2018 to $900 in 2026. Under Armour's HeatGear shirt went from $27.99 in 2015 to $30 in 2023, and was on sale at $14.97 in 2024.

    Arc'teryx 2018 ↗ · Arc'teryx 2026 ↗ · Under Armour 2015 ↗ · Under Armour 2024 ↗ · Sweetgreen results 2025 ↗ · Sweetgreen $10 bowl ↗

  9. 09

    Decide how you get paid and what you own.

    A pattern from four hardware cases. In hardware, how you charge and what a rival can't copy, like patents, data or a community, can matter as much as the product. Humane asked $699 for its AI Pin plus a required $24 a month, before the product had earned it. It never broke out.

    ExampleUltrahuman sold its Ring Air for $349 with no subscription, the same price as Oura's ring, which charges a monthly membership. Its sales grew almost six times in 2024, to $74.5m. Then Oura won a patent case at the US International Trade Commission, and a US import ban from 21 Oct 2025 cut Ultrahuman's US share from about a quarter of the market to low single digits. Strava won on how it charges. In 2019 its co-founder came back as CEO and moved the team from about 20% working on what subscribers pay for to over 90%. In May 2020 it put its best features behind one subscription. Sales grew about 70% to about £170m in 2021.

    TechCrunch, Ring Air ↗ · Entrackr ↗ · USITC ↗ · TechCrunch, US share ↗ · Goldman Sachs Talks ↗ · Guardian ↗ · CNBC, Humane ↗

  10. 10

    Watch profit and company value as closely as sales.

    In brands that rose and fell, company value fell a median 86% from its high (9 brands, falls of 79% to 97%). 8 of the 12 with a profit high went from an operating profit to a loss. Sales often held up for longer: 6 of 15 were at their sales high at the latest count.

    ExampleGoPro's shares reached $98.47 in 2014 and were down to $9.80 by November 2016.

    GoPro 10-K 2014 ↗ · Bloomberg ↗

  11. 11

    How big to bet is still a judgement call.

    Winners put a median 4.5% of what they had on the bet (15 bets). Bets that didn't win put 19.5% (8 bets). Smaller still looks wiser, but less clearly than before, and that is too few bets to call it a rule.

    ExampleIn 1997 Steve Jobs cut Apple down to a few products. It spent $217m on restructuring, about 15% of its cash and investments, and staff fell from 8,437 to 6,658. Apple went from a $1,045m loss in 1997 to a $309m profit in 1998. A big slice, and it won big.

    Apple 10-K 1997 ↗ · Apple 10-K 1998 ↗

How to use it

Before a big move, sit down with the people making the call and answer these straight.

  1. Who doubts this, and what exactly are they missing?
  2. Is this a new category, or a better version of something that already exists?
  3. What are we saying no to, and what does that make room for?
  4. Does this break anything we've promised our customers?
  5. Can we fund two years of this without needing it to work early?
  6. Are any red flags up? Scaling before proof, drifting off our core, money that can leave, one key person with no plan B.
  7. Are we ready for this to take years, and to look quiet while it works?
  8. If it works, how do we keep telling the story, grow in steps and earn a higher price?
  9. How do we get paid, what do we own that a rival can't copy, and do we control how we sell?
  10. How much of what we have are we putting on this, and why that much?

What we're still testing

  • How big to bet.4.5% against 19.5% looks like a lesson, but it rests on 23 bets, and two of Steve Jobs's biggest 1997 bets, 22% and 15% of Apple's cash, won big. Frank Lloyd Wright's Fallingwater ran to about four times its estimate, and it won big too.
  • Killing your own darling.These bets won 56% of the time (25 bets), and 7 of 15 on the fully checked ones. Steve Jobs's 1997 cuts lifted it. It seems to work when the next thing is already clear.
  • Building a community or a culture first.The numbers look good, 89% won (47 bets), and 10 of the 11 fully checked ones won. It is very close. It stays off the list until it holds on a bigger fully checked group.
  • Keeping control of how you sell.Bets that own the channel won 74% of the time (43 bets). The newest brands in the study point the same way. A pattern to watch next to rule 9.
  • How you get paid.Rule 9 rests on four cases so far.
  • The price after a stall.Of the four stalled brands we have prices for, three cut the hero price and one barely moved it. Still a small group.
  • How the study is built.We choose the bets, and some tags are our own judgement after the fact. Most of the brands broke out, which can flatter a pattern. A blind test on 10 new brands is next.

Updated weekly

Every week we add new bets and new brands, check more of the old ones and run the numbers again. Last updated 18 Nov 2026 (v1.2).

  • 18 Nov 2026A steady week. Doubted bets edged up, 81% won against 31%. Saying no is 11 of 11 on the fully checked bets. Bets with no red flag won 77%. Two more brands cut their hero price once growth stalled, so rule 8 is firmer. Rule 9 has a fourth case. Fully checked bets went from 94 to 104.
  • 4 Nov 2026Doubted bets moved further ahead, 80% won against 32%. Red flags got sharper, 27% against 76%, and taking the same playbook to a new country joined the warning signs. Saying no is 9 of 9 on the fully checked bets. Betting small looks less clear-cut, and rule 11 now has an example. Fully checked bets went from 79 to 94.
  • 28 Oct 2026Doubted bets pulled further ahead, 78% won against 32%. Red flags got sharper. Saying no had its first outright loss. Rule 9, how you get paid and what you own, is new. Fully checked bets went from 69 to 79.
  • 21 Oct 2026Saying no and keeping your promises became rules of their own. The quiet breakout joined rule 7. Building a community moved to the testing list.
  • 14 Oct 2026First plain-English version, with 9 rules.

Got a big move on the table?

We use this with founders and their teams to shape the bet and make it land.

→  See how we work