The Flywheel

The cases

1. 2006

   Mastercard

   Held

   Whether to convert the bank-owned cooperative into a public company via IPO

   • Go public via IPO — Convert the cooperative into a public company owned by shareholders.chose this
   • Remain a bank cooperative — Keep the network owned by its member financial institutions.

   Stated reasonMembers voted overwhelmingly in favor of the offering, framed as a modernization of the network's structure

   What happened

   • 2024 Net revenue: net revenue rose 12% to $28.2 billion in FY2024
   • 2024 Switched transactions: 159.4 billion switched transactions in 2024, up 11%
   • 2023 Merchant antitrust settlement: a $5.54 billion settlement was approved in 2019 and affirmed on appeal in 2023
   Read the full analysis →
2. 2014

   Progressive

   Held

   Should Snapshot only reward safe drivers, or also surcharge risky ones?

   • Discount only — Reward safe drivers, never raise prices on risky ones.
   • Price risk both ways — Reward safe drivers and surcharge risky ones.chose this

   Stated reasonSurcharging risky drivers prevents adverse selection, keeping the pool selected while rivals' pools drift toward losses.

   What happened

   • 2023 continuous monitoring reach: live for new customers in 30 states, representing about half of personal auto premiums written
   • 2023 combined ratio and net premiums written: combined ratio of 94.9 while growing net premiums written 20% to $61.6 billion
   • 2025 Personal Lines share of premiums: Personal Lines drove 87% of net premiums written
   Read the full analysis →
3. 2018

   DoorDash

   Held

   Whether to give away per-order delivery margin through a paid subscription to buy frequency and retention

   • Launch paid subscription giveaway — Waive delivery fees for members to grow frequency and retentionchose this
   • Keep charging full per-order fees — Preserve per-order margin on every delivery

   Stated reasonMembers retain better and order more frequently than non-members, so surrendering per-order margin buys frequency and retention

   What happened

   • 2024-12-31 DashPass and Wolt+ members: over 22 million members
   • 2024 Full-year GAAP net income: net income $123M vs. net loss $(558)M in 2023; first full-year GAAP profit
   • 2024 U.S. food delivery market share: roughly 60.7%, more than double Uber Eats at 26.1%
   Read the full analysis →
4. 1956

   McDonald's

   Held

   Whether to remain a thin royalty collector or acquire the land under each franchise and sublease it back at a markup

   • Collect royalty alone — License the name for 1.9% of franchisee sales
   • Own the land and sublease — Control the real estate and sublease to franchisees at a markupchose this

   Stated reasonThe 1.9% royalty was too thin to fund expansion; owning and subleasing land produced a stable cash stream and appreciating collateral

   What happened

   • 1960 system gross sales and locations: restaurants grossing roughly $56 million a year across some 200-plus locations
   • 2025 real-estate holdings on balance sheet: property holdings sit at $37.7 billion
   • 2025 franchised share of restaurants: about 95% franchised
   Read the full analysis →
5. 2001

   Tesco

   Held

   Whether to keep Dunnhumby as a vendor or buy the analytics firm behind Clubcard

   • Keep Dunnhumby as vendor — Retain analytics via arm's-length contract and renewal
   • Buy the partner — Take a controlling stake in Dunnhumby to own the data enginechose this

   Stated reasonOwning the firm that turns swipes into decisions converts a marketing program into a compounding information asset rivals cannot replicate

   Committeda reported £30m (for the initial 53% stake in Dunnhumby in 2001)

   What happened

   • 2025-02 Clubcard households: more than 23 million UK households now carry a Clubcard - roughly four in five homes in Britain
   • 2024 Clubcard app UK users: a Clubcard app carrying 12.7 million UK users feeding a retail-media and insight business
   • 2010 ownership of Dunnhumby: raised stake to 84% in 2006 and completed full ownership by roughly 2010
   Read the full analysis →
6. 2011

   Alibaba

   Held

   Whether to keep Alipay inside Alibaba Group or separate it out under founder control to clear the regulatory path

   • Keep Alipay inside group — Retain ownership of the payment ledger, data and compounding credit engine
   • Separate Alipay to founders — Transfer Alipay to Ma and co-founders to secure a clean payment licensechose this

   Stated reasonThe transfer was justified as regulatory: a payment license might be denied to companies tangled in foreign-investor VIE structures, so Alipay had to be made domestically clean

   What happened

   • 2020 Ant Group IPO: suspended by regulators days before it priced
   • 2023-07 Ant Group fine: fined roughly $985 million for compliance failures
   • 2011 Yahoo's stake value in Alipay: reduced by more than 60%
   Read the full analysis →
7. 2014

   Progressive

   Held

   Whether to use telematics data only to discount safe drivers or to also surcharge risky ones asymmetrically

   • Discount safe drivers only — Use telematics purely to offer renewal discounts
   • Price asymmetrically both ways — Discount safe drivers and surcharge risky oneschose this

   Stated reasonTelematics is the most predictive rating variable, so pricing both directions selects better risk

   What happened

   • 2024 combined ratio: 88.8 combined ratio, more than 7 points under its 96 target
   • 2024 net income: net income of roughly $8.5 billion — more than double 2023's $3.9 billion
   • 2024 direct private-auto premiums: lifted direct private-auto premiums 24.5% to $60.05 billion in a single year
   Read the full analysis →
8. 2009

   Domino's

   Held

   Whether to publicly broadcast customer criticism and relaunch a redesigned recipe

   • Publicly self-critique and relaunch — Air customer complaints and redesign the recipe openlychose this
   • Keep quiet, iterate privately — Improve product without broadcasting criticism

   Stated reasonThe recipe had genuinely gotten worse and the company chose to admit it and relaunch a redesigned product

   Committed$75 million (six-week marketing campaign paired with employee retraining)

   What happened

   • 2018 consecutive same-store sales growth: By 2018 Domino's had clocked its 30th consecutive quarter of same-store sales growth
   • 2026 15-year total return: The fifteen-year total return ran to something like 1,940%
   Read the full analysis →
9. 2009

   Shopify

   Held

   Whether to build the merchant base first and open the developer App Store only after a market existed

   • Launch App Store later — Build merchant base first, open developer ecosystem afterwardchose this
   • Launch App Store at start — Offer developer platform as an original core feature

   Stated reasonBuild the merchant base first, then invite developers to build on a market that already existed

   What happened

   • 2024 GMV through platform: platform facilitated GMV of $292.3 billion in 2024 (up 24% YoY)
   • 2024 payments penetration: Shopify Payments penetration reached 61.9% of GMV in 2024, with $181.0 billion processed
   Read the full analysis →
10. 2015

    PayPal

    Held

    Whether to keep routing ~80% of eBay's sales through PayPal after the spinoff

    • Contractual routing lock — eBay keeps ~80% of GMV on PayPal for five years with penaltieschose this
    • Clean operational break — Both companies fully independent to compete at spinoff

    Stated reasonThe flywheel was too valuable to let stop spinning while everyone adjusted

    What happened

    • 2020 Routing dependence duration: Stayed structurally dependent on eBay traffic until 2020
    • 2024 Total payment volume: $1.68 trillion in total payment volume in 2024
    Read the full analysis →
11. 2020

    Airbnb

    Held

    In the guest-vs-host conflict, protect guest trust or host trust?

    • Protect guest trust — Override host terms and refund guests in fullchose this
    • Protect host trust — Honor host cancellation policies and payouts

    Stated reasonThe cycle runs on guest trust; a guest burned by one bad stay distrusts the next ten

    Committed$250 million relief fund (paid hosts roughly 25 cents on the dollar of normal cancellation payouts)

    What happened

    • 2024 active listings: crossed 8 million active listings
    • 2023 host earnings: hosts earned more than $57 billion in 2023 alone
    Read the full analysis →
12. 2020

    TikTok

    Reversed

    How to structure creator pay: a fixed pool split among all, or a performance-weighted incentive

    • Fixed-pool Creator Fund — A fixed pool of $200 million split among all eligible creators.chose this
    • Performance-weighted rewards — Pay weighted to performance and longer content.

    Stated reasonThe interest-graph feed made any creator interchangeable, so TikTok never needed to pay a premium to keep stars.

    Committed$200 million (Starting pool of the US Creator Fund.)

    What happened

    • 2023 Creator Fund shutdown: TikTok shut the fund down in late 2023 and swapped in the Creator Rewards Program
    • 2024 revenue: an estimated $23 billion in revenue in 2024
    Read the full analysis →