The decision

Build, defend or charge for the thing that keeps customers from leaving.

What the record shows

24 of 31 judged decisions were never undone. In all but one walk-back, the first move was the company's own.

The lesson

A moat is rarely stormed. It is drained by its owner, usually through a new price or a looser standard.

The fee Unity charged for its own success

On September 12, 2023, Unity, which makes software that many video games are built on, announced a new fee: a charge every time a game made with it was installed. The point was to collect more from the hit games it most needed to keep.

Ten days later Unity took the fee off its free tier and spared games built on older versions, with a public apology. Within a month its chief executive had retired. On September 12, 2024, a year to the day, Unity canceled the fee entirely and went back to charging per seat, with its Pro tier raised 8% to $2,200 a year.

Developers didn't leave all at once. In 2024, 51% of the games released on Steam still used Unity. But by revenue, games made with its rival Unreal took 31% of Steam sales against Unity's 26%, the first time Unreal had led since 2018.

That is the whole decision in one story. Unity's advantage was never one piece of code. It was every game already built on the engine and every developer who knew it. The fee tried to charge for that advantage, and in doing so put it up for grabs. The rest of this guide is about the companies that built, defended or spent an advantage like it, and what happened next.

What moat is

Definition

A moat is whatever keeps customers from leaving and rivals from copying you, even after everyone can see how you win. The decision is how to build one, how to defend it, and how much to charge for it.

Moats come in a few kinds. Some are networks that get more valuable as they grow, like a card network that every bank belongs to. Some are habits and skills, like the code millions of developers have already written for one company's chips. Some are brands, some are legal rights, and some are ownership structures that put control out of a bidder's reach. Each can be built, defended or spent, and the record has examples of all three.

Why it is so hard

On paper, protecting an advantage sounds like the easiest decision a company makes. In a real company, three things make it hard.

1

Nobody owns it

Revenue, margin and growth each have an owner with a target. The reason customers stay usually has no line in the plan, so it loses arguments on points.

2

It looks like pricing power

Customers who can't easily leave look like customers who will pay more. That is true up to a point, and nobody can see the point in advance.

3

Every test is public

A moat is only tested when someone pushes on it: a rival, a bidder, a court or the company itself. By then the push is in the open, and backing down is expensive.

The ways companies did it

The decisions on record take four shapes. Two make the advantage harder to take; two find out what it will bear.

Build it

Invest in what a rival can't copy quickly

Spend years on something a competitor would need years to match.

Nvidia
Introduced CUDA in November 2006, a platform for programming its chips for general computing. By its 2025 filing over 5.9 million developers used CUDA and its other tools, and its share of AI training exceeded 90%.
Eli Lilly
Began building factories for its drug tirzepatide in 2020, before any approval, on the view that capacity can't be built on a rival's timeline. Its U.S. manufacturing commitment since 2020 passed $50 billion, and by August 2024 every dose was available.
Lock it

Put control out of reach

Build the advantage into who owns what, so no rival or bidder can simply buy it.

Visa
In June 1970 took its card program out of Bank of America's hands and reincorporated it as a company owned by the issuing banks together, turning rivals into co-owners. In fiscal 2024 it processed 234 billion transactions on its own rails.
Hershey
In 2016 its board unanimously rejected a roughly $23 billion bid from Mondelēz. A change of control would have handed its Kit Kat license back to Nestlé for free.
Defend it in court

Make the law hold the wall

Use rules, patents or contracts to keep rivals out, and go to court to test them.

American Express
Refused to settle the government's case against its rules stopping merchants from steering customers to cheaper cards, as Visa and Mastercard had. In 2018 the Supreme Court ruled for it, 5-4. In 2024 it earned a record $10.13 billion on $65.95 billion of revenue.
Moderna
Sued Pfizer and BioNTech in August 2022 over three of its mRNA patents. The case was stayed in April 2024, and in March 2025 a patent board ruled every challenged claim in two of the three patents unpatentable.
Price it

Find out what the advantage will bear

Raise what customers pay, or ration what they can get.

Ferrari
Treats supply as a dial, keeping demand ahead of it. In 2024 it shipped 13,752 cars at a 38.3% EBITDA margin, and cut shipments to mainland China, Hong Kong and Taiwan by 328 cars to protect exclusivity.
Delta Air Lines
In September 2023 tied SkyMiles elite status mainly to credit-card spending. Its chief executive said the changes "probably went too far", and Delta walked key provisions back within months.

What 35 companies did

Here is every decision on record, placed by the year it was made and marked by how it turned out. Hover a mark, or open the table, for the company and the outcome.

35 defending an advantage decisions, 1891 to 2026

Each mark is one decision. "Never undone" means the company did not reverse it, not that it worked.

Never undone24Kept in part5Walked back2Too early to call4
1890191019301950197019902010
Show as a table
CompanyYearChoiceOutcome
Coca-Cola1891Keep perpetual trade secretNever undone
Caterpillar1925Independent dealer networkNever undone
Visa1970Issuer-owned cooperativeNever undone
Red Bull1984Sell meaning at premiumNever undone
Diageo1997Own the whole barNever undone
Crocs2006Acquire Jibbitz charm businessNever undone
Nvidia2006Fund a developer platformNever undone
BMW2009Run the 'Joy' campaignWalked back
PepsiCo2010Direct-store-deliveryNever undone
Apple2011Sue on design patentsKept in part
Diageo2012Build the aging reserveKept in part
Etsy2013Allow outside manufacturersKept in part
Ferrari2014Managed scarcity dialNever undone
Hershey2016Reject Mondelēz bidNever undone
Palantir2016Sue the ArmyNever undone
Crocs2017Cut to one hero clogNever undone
American Express2018Litigate to Supreme CourtNever undone
Diageo2018Seek SWA permissionNever undone
L'Oreal2018Acquire ModiFace outrightToo early to call
Novo Nordisk2018Grow via GLP-1 moleculeNever undone
AMD2019Split compute and I/ONever undone
Airbnb2020Suspend performance marketingNever undone
Eli Lilly2020Build capacity at riskNever undone
LVMH2021Close at renegotiated priceNever undone
Moderna2022Sue Pfizer and BioNTechNever undone
Nvidia vs AMD2022Fund CUDA compatibility layerNever undone
Adobe2023Abandon the dealNever undone
Delta Air Lines2023Overhaul SkyMiles toward card spendKept in part
Unity2023Per-install runtime feeWalked back
Ralph Lauren2024Exit discount wholesaleNever undone
Tesla2024Open the networkNever undone
Universal Music2024Pull catalog from TikTokKept in part
Airbnb2025Impose mandatory 15.5% feeToo early to call
Roku2026Sell to FoxToo early to call
Stripe2026Bid to acquire PayPalToo early to call
Source: Stratrix Decision Records, Moat Anatomy (35 verified records). Outcome as recorded in each case.

Of the 31 decisions that can be judged, 24 were never undone, 2 were walked back and 5 were kept only in part.

A few decisions are filed here but are really about something else, such as LVMH's 2021 purchase of Tiffany. Read them as neighbors of this decision.

The walk-backs

Put the walk-backs side by side and one thing stands out. Only one began with a rival pushing in: Apple's lawsuit over phones it said Samsung had copied. Every other one began with something the company did to its own advantage.

CompanyWhat it didHow it was undoneKind of move
BMW2009Replaced The Ultimate Driving Machine with a Joy campaign in North AmericaBrought the old tagline back by 2012Swapped the promise
Apple2011Sued Samsung over three design patents covering the iPhone's lookWon $1.049 billion from a jury in 2012; the Supreme Court reversed the design-patent damages 8-0 in 2016, and the case settled in 2018 on undisclosed termsLeaned on a legal wall
Diageo2012Committed roughly £1 billion to expand its Scotch capacity and build new warehousingIn September 2025 temporarily halted production at Balcones, George Dickel and TeaninichLeverage of 3.4x in fiscal 2025, outside its own target rangeBuilt ahead of demand
Etsy2013Let sellers use outside manufacturers so the marketplace could scaleBy 2024 was banning 3.5 million spam accounts and cutting off tools for listing mass-produced goodsMarketplace sales of $10.5 billion in 2025, 14% below the 2021 peakLoosened the standard
Unity2023Charged developers a fee per game installCanceled it in September 2024 and went back to charging per seatUnreal passed Unity in Steam game revenue, 31% to 26%Changed what customers pay for
Delta Air Lines2023Tied SkyMiles elite status mainly to credit-card spendingWalked key provisions back within monthsChanged what customers pay for
Universal Music2024Pulled its whole catalog from TikTok to force better termsCame back in May 2024 with a deal whose financial terms were not disclosedTested its leverage

The kind of move is Stratrix's reading of each case. Apple, Diageo, Etsy, Delta and Universal Music were kept only in part; BMW's and Unity's were undone.

In five of the rows the first move was a push on the company's own advantage: a new tagline, a looser standard, a new fee, a new rule for status, a blackout. The two that changed what customers pay for, Unity and Delta, were both pulled back within a year.

The other two show the limits of building and defending. Diageo locked capital into barrels on the bet that Scotch demand would keep climbing; its net sales in fiscal 2025 were flat at $20.2 billion, and in September 2025 it paused three distilleries. Apple won its case in front of a jury, and then the Supreme Court reversed the design-patent damages 8-0.

BMW's row is the strangest. Its North American sales rose during the Joy campaign, from 195,502 cars in 2009 to 220,113 in 2010, and it brought The Ultimate Driving Machine back by 2012 anyway. A good year was not enough to make the new promise worth more than the old one.

A moat is rarely stormed. It is usually drained by the company that owns it.

How to run this decision

If you are weighing a version of this choice, here is what the record suggests doing, in order.

  1. Write down why customers stay, in one sentence

    Not the feature list and not the switching cost: the reason. If it takes a paragraph, nobody in the company will recognize an attack on it, including one from inside. BMW's reason was a promise about how its cars drive, and it set that promise aside for three years.

    Seen in: BMW, 2009
  2. Name the part that is not for sale

    Decide in advance which piece of the advantage stays off the table whatever the revenue case says, and write it where the people setting prices can see it. Etsy let sellers use outside manufacturers in 2013 to grow, and by 2024 was cutting off tools that let sellers list mass-produced goods.

    Seen in: Etsy, 2013
  3. Treat a change in what customers pay for as a product change

    Customers react to a new unit of price far more than to a higher number. Unity switched from charging per seat to charging per install, and Delta switched elite status from flying to card spending. Both pulled back within a year.

    Seen in: Unity, 2023 · Delta Air Lines, 2023
  4. Build what a rival can't compress

    The strongest advantages here took years to build and would take a rival years to copy: code written by developers, factories built and approved. Eli Lilly began building capacity for tirzepatide in 2020, before any approval, because capacity can't be built on a competitor's timeline.

    Seen in: Eli Lilly, 2020 · Nvidia, 2006
  5. Test a legal wall before you lean on it

    Patents and contract rules hold only until a court looks at them. American Express defended its merchant rules all the way to the Supreme Court and won, 5-4. Moderna sued over its mRNA patents, and two of the three had every challenged claim ruled unpatentable. Apple won $1.049 billion from a jury over the iPhone's design, and the Supreme Court later reversed the design-patent damages.

    Seen in: American Express, 2018 · Moderna, 2022 · Apple, 2011
  6. Decide what a refusal looks like before you launch

    Before a change that pushes on customers, define the reaction that would make you stop and the date you will read it. Unity revised its fee ten days after announcing it and canceled it a year later; deciding in advance would have been cheaper than deciding in public.

    Seen in: Unity, 2023
  7. Know what a sale would do to the advantage

    If part of the advantage sits in a contract or a structure, find out what a change of control would do to it before a bidder asks. Hershey's Kit Kat license would have gone back to Nestlé for free under new ownership, and its board unanimously rejected a roughly $23 billion bid from Mondelēz in 2016.

    Seen in: Hershey, 2016 · Visa, 1970

To work the numbers for your own case, read Competitive moat analysis: how to test whether your advantage is real.

Warning signs

What it has looked like when this decision was going wrong.

  • The proposal changes what customers pay for, not just how much.Unity, 2023 · Delta Air Lines, 2023
  • A long-running brand promise is being replaced because leadership has a better idea.BMW, 2009
  • The standard that defines the product is relaxed so the network can grow faster.Etsy, 2013
  • The advantage rests on patents nobody has tested in court.Moderna, 2022
  • The plan assumes a partner needs you more than you need it.Universal Music, 2024

All 35 cases

Each case is one company at one decision: what was on the table, what it chose, why, and what followed, with every fact sourced.

Never undone · 24

Kept in part · 5

Walked back · 2

Too early to call · 4

Questions

What is a moat in business?
Whatever keeps customers from leaving and rivals from copying a company, even when its strategy is plain to see. Networks, developer habits, brands, legal rights and ownership structures are all moats in this guide.
What are examples of companies with strong moats?
Visa, whose 1970 structure made the issuing banks co-owners of its network; Nvidia, with over 5.9 million developers using CUDA; Caterpillar, with its network of independent dealers; and Ferrari, which keeps demand for its cars ahead of supply on purpose.
How do companies lose their moats?
In this guide, mostly by their own hand. Most of the decisions that were walked back began with the company changing what customers paid for, relaxing its own standard, pulling its catalog from a partner, or replacing its own brand promise.
How do you build a moat?
By investing early in something a rival cannot copy quickly: a developer platform, manufacturing capacity, a dealer network or an ownership structure. The examples on record took years to show up in results.

More from Stratrix

Run this decision
The Moat Anatomy

“What would it actually cost a challenger to take your customer?”

See the pack — $499 →

Ask the Analyst

Three questions a month are free.

All Access

Every Casebook, three Strategy Packs a seat each year and the Analyst, for teams that face these decisions all year.

$1,999 a year · $499 a seatWhat is included →

New analyses, by email

The defining moves in business, checked against the record, as they publish.