In the long run, you get what you incentivize
I didn’t come up with this. But it took me a long time to really digest it. It applies to business, sure. It also applies to policy design, team management, education, even how you relate to yourself. Any system, run long enough, ends up looking exactly like whatever its incentive structure rewards.
The flip side is harsher: if your goals run against people’s long-term incentives, then no matter how strong the system, how strict the rules, how deep the pockets, you’re just manufacturing friction. Brute force might hold things together for a while. But over time, you will always lose to structures that work with human nature, not against it.
I only understood later why Buffett and Munger treat business models as the core thing to examine in an investment. It’s not because a model tells you “can this company make money” (plenty of bad business models make money in the short run). It’s because it tells you something far more important: to make that money, what is this company being trained to become?
That’s the real question.
A business model is a behavioral training machine
Most people, when they talk about business models, talk about revenue streams, margins, pricing strategies. That’s all real, but it’s all surface.
Go deeper, and a business model is an organization-level incentive structure. It determines what behavior gets rewarded most easily inside the company. And that reward signal, day after day, shapes the organization’s habits.
Picture two companies.
The first one makes most of its profit from long-term user renewals. Its sales, product, and support teams wake up every morning thinking: “How do we make users feel this is worth it, so they stay?” Nobody needs to paint that on the wall. It’s just the path of least resistance to making money.
The second one makes most of its profit from first-time purchases by new customers. Organizational energy naturally flows toward marketing, messaging, channels, packaging. Product matters too, but it’s treated more as material for closing that first sale. What happens after the user buys? Doesn’t directly move the number that matters most on the P&L.
Both companies can be profitable. Short-term, the second one might even make more. But give it time, and they become completely different organizations. One keeps getting better at delivering value. The other keeps getting better at selling the promise.
A business model doesn’t describe how a company makes money. It describes what kind of organization the company is being trained to become, to make that money.
Munger said something like: “Show me the incentive and I’ll show you the outcome.” I’d add: show me what a company looks like after five years, and I can reverse-engineer what its business model was rewarding.
Costco: when “delighting end customers” becomes the easiest way to make money
If you’re not familiar with Costco, here’s the short version.
Costco is an American warehouse-style membership retailer. You pay an annual fee (65 dollars for a basic membership, 130 for the executive tier) before you can shop there. Once inside, everything is priced aggressively low, because Costco caps its gross margins at roughly 14%. For comparison, a typical supermarket runs 25% to 35%.
In other words, Costco’s markup barely covers operating costs. The company’s real profit comes from membership fees.
This simple structural tweak has consequences that go far deeper than most people realize.
First, the profit anchor shifts. When your main profit source moves from “how much we make per item” to “how many people renew their membership,” the entire organization’s behavioral direction gets rewritten. The number Costco cares about most is renewal rate (currently stable above 90%). To get nine out of ten users to voluntarily pay again every year, the only sustainable approach is to actually make them feel it’s worth it.
Second, low margins become a self-training device. Because you can’t coast on fat markups, Costco is forced to keep grinding on supply chain efficiency, selection discipline, inventory turnover. It carries about 4,000 SKUs (a typical supermarket stocks 30,000 to 50,000), with only a few curated options per category. This is the sharpness that comes from self-discipline: do the filtering for the customer, stake your reputation on it.
Third, Costco’s way of pleasing users is dead simple, which is exactly why it lasts. Lots of companies want to delight users. But their definition of “delight” is too broad: subsidies, gimmicks, over-the-top service, emotional marketing. All of it is expensive, hard to replicate, unsustainable. It becomes a bottomless pit. Costco doesn’t play that game. It does three things: low prices, reliable quality, consistent experience. Every time you walk into Costco, you can verify it’s not lying to you. Prices really are lower. Stuff really is solid. Repeat that verification a hundred times, and it turns into trust. Many companies spend fortunes trying to “manufacture” trust. Costco just lets you “confirm” it yourself.
Fourth, and the one I think matters most: the exit door is always open. Costco members can stop renewing at any time. No multi-year contracts. No cancellation fees. No customer service maze that makes you call three times before they let you quit. The user’s veto power is never taken away.
Every year, Costco faces a popular referendum: are you still worth my money?
That’s Costco’s real moat. Users can leave, but they don’t want to. On the surface it looks like “no barriers.” In reality, precisely because it never relies on force to keep people, it’s compelled to keep making “delight end customers” the easiest way to earn.
These four layers interlock. Profit comes from renewals, so you must make users feel it’s worth it. Margins are compressed, so you must win on fundamentals. The way you please users is simple, so trust accumulates. The door is always open, so you never dare slack off. Four expressions of the same incentive structure.
Counter-examples: when incentive structures reward the wrong behavior
Understanding Costco makes it easier to see two industries where incentives have gone badly wrong.
US health insurance: profit comes from denying claims
The American health insurance industry is the textbook case of incentive misalignment.
The business model is straightforward: collect premiums, pay out as little as possible. You pay thousands, sometimes tens of thousands of dollars per year in premiums. But when you actually get sick and need the money, the insurer’s profit depends on how many claims it can reject.
So the most profitable capability in the entire industry isn’t making you healthier. It’s claims denial management. Insurers employ armies of people to review claims, find rejection rationales, and design terms so complex that when you need help the most, you can’t even understand your own rights. The US spends over $1,000 per person per year just on healthcare administration, far more than any other developed country. A big chunk of that goes toward figuring out how to legally not pay you.
The payer and the beneficiary are the same person, yet the incentives point in opposite directions: you paid, hoping to be taken care of when sick; the company collected, hoping you don’t show up when you are.
At an industry level, the entire profit mechanism systematically rewards “not helping the customer.” Within this structure, an insurer that genuinely, eagerly pays claims will have worse margins than its peers. It gets punished competitively. The incentive structure penalizes the right behavior.
Enterprise software: profit comes from you being unable to leave
Oracle and SAP represent perhaps the most refined version of the “lock-in moat” in traditional enterprise software.
Massive implementation costs (a large ERP project can easily run into millions of dollars and take years). Deep customization embedded into business processes. Systems so complex they require specialized consultants just to maintain. And periodic license audits, which are, at bottom, compliance intimidation: you may have unknowingly violated licensing terms, and after the audit, you owe a hefty sum.
Once a customer is on the system, switching is nearly impossible. Not because the system is great, but because the cost of leaving is too high. So the easiest way to make money becomes raising switching costs, not improving the product.
SaaS was supposed to be the rebellion against this incentive structure. The early philosophy was pure: pay monthly, no long-term lock-in, users can leave anytime, the company is forced to prove its worth every month. That logic is identical to Costco’s.
But here’s the irony. Many SaaS companies, once they scaled, walked right back down the old path. Take Salesforce. It now pushes 3-to-5-year contracts. Early cancellation means paying the full remaining balance. Auto-renewal clauses catch small businesses off guard. Data migration costs are high enough that people just don’t bother. The original “leave anytime” promise turned into a different form of lock-in.
The gravitational pull of incentive structures is real. When “profiting from friction” is easier than “profiting from value,” even companies that started under the banner of “anti-lock-in” get pulled back. Very few companies can resist this gravity. That’s what makes Costco’s design worth studying: it structurally blocks the “profit from friction” path. The company never even gets the chance to try.
The curse of scale
Beyond model-level misalignment, there’s a subtler problem: even a company that starts with the right incentive structure can lose it after getting big enough.
There’s an old Chinese saying: when the shop is big, it bullies the customer; when the customer is big, they bully the shop. Same thing happens today, when a company grows large enough that users’ veto power weakens, it starts testing boundaries without even realizing it.
In early startup days, you genuinely need each specific user. When that user says “I’m unhappy, I’ll leave,” you lose sleep over it. You naturally see yourself as the one who must earn their business. You do everything to serve every single person.
But once you’re big enough, once users are plentiful, a subtle mental shift creeps in. Losing one user doesn’t matter anymore, because there are always more coming. You’re confident in your growth, so you stop caring about each individual. A user is unhappy? The aggregate numbers look fine. Support is slow? Doesn’t affect the quarterly report. Terms get complicated? Most people won’t read them anyway.
This isn’t a moral failing. It’s the incentive structure at work. When users’ veto power weakens (fewer alternatives, higher switching costs, deeper ecosystem lock-in), the short-term cost of mistreating them goes down. The organization naturally drifts in that direction.
This drift shows up in three ways:
Choice dilution. Users have fewer alternatives. Switching costs rise. Even when unhappy, they can’t be bothered to switch. The company feels less pressure.
Attention dilution. Users go from “a specific person” to “a data point in DAU.” “Delighting end customers” shifts from an instinctive reaction to something that needs to be written into OKRs before anyone acts on it.
Principal dilution. The company starts serving many masters simultaneously: users, shareholders, regulators, partners, internal processes. It talks about “customer obsession,” but when quarter-end revenue targets conflict with user experience, user experience never wins.
So “earning it every day” can’t just be early-stage startup passion. It needs to become institutional. It needs to be translated into one thing: keeping the user’s veto power permanently real.
When that veto power exists, the company has no choice but to stay humble. When it disappears, humility can only be sustained through the founder’s personal discipline. And discipline that depends on one person doesn’t scale.
The most dangerous paradox: moats create learned helplessness
Now pull the threads together, and you see a deeply ironic paradox.
Many companies work desperately to build moats that lock users in. But it’s precisely this lock-in that pushes users into learned helplessness.
There’s a classic psychology experiment: dogs are repeatedly shocked but unable to escape. Eventually, even when the cage door opens, they stop trying to leave. Users locked in by moats work the same way. It’s not that they don’t want to leave. They tried, found they couldn’t, and eventually gave up. The desire to compare fades. The motivation to give feedback fades. The courage to leave fades. What’s left is numb renewals and silent complaints.
Think about whether you’ve experienced this yourself: you’ve used some software for years, you’ve been unhappy with it for a long time, but your data is all on it, your workflows are built around it, switching means relearning everything. So you just keep putting up with it. You’re still paying, but you’re no longer an “active user.” You’re just a “trapped payer.”
When users become this, the company changes too. Because the biggest source of external pressure, “users might leave,” has vanished. Without external pressure, organizations slowly calcify. The drive to build new things weakens, because existing users can’t leave anyway. Service standards drop, because complainers have no alternatives. The smartest people get reassigned to retention mechanics and upselling, not to building the product.
The moat looks deeper. In reality, the company has eliminated the external force that used to push it to improve. It held onto its existing base, but hollowed itself out in the process.
So moats come in two very different types.
Lock-in moats: rely on friction, switching costs, and cancellation difficulty to keep users. Users stay because leaving is too much trouble. Looks good on the financials short-term. Long-term, it dulls the company. It stops obsessing over delighting end customers, because the structure rewards “making it harder for users to leave.”
Earned moats: the door is always open. Users can walk out anytime. Retention comes from sustained value and trust. Every time users consider leaving, they decide it’s still worth staying. So they choose to stay.
Costco is an earned moat. Users can skip renewal every year. Nothing prevents them from leaving. But precisely because of that “leave anytime” design, Costco’s organizational culture is forced into continuously making users feel it’s worth staying. A 90%-plus renewal rate wasn’t locked in. It was earned.
Sustainable altruism
One last thing. A personal belief I’ve gradually formed.
Good business is, in the end, sustainable altruism.
That might sound too idealistic. But I’m not saying “good companies should do charity.” I’m making a more practical observation: if a company’s way of making money structurally requires it to hurt users (raising switching costs, creating information asymmetry, profiting from user inertia or confusion), it’s short-lived by nature. You may be simultaneously increasing both the user’s costs and the user’s pain. That bill doesn’t show up immediately, but over time, it always comes due.
The reverse holds too. The best business models are ones where the company’s structure naturally points toward delighting end customers, where that delight becomes the simplest path to profit. None of that depends on the founder being a good person or on culture slogans on the wall. It depends on incentive mechanisms that make “doing the right thing” the default, and “taking advantage of users” genuinely difficult.
Call it pragmatism more than morality: structures that make money by working with human nature last longer than structures that make money by working against it.
Profitability, moats, and user satisfaction often aren’t three separate things. They’re three manifestations of the same thing across time. As long as “delighting end customers” is the path of least resistance to profit, the organization accumulates trust, process discipline, and scale efficiency over the long run. None of that can be copied overnight. The moat isn’t dug from the outside. It’s walked into existence, day by day, along the easiest correct path.
When big companies start mistreating customers, it’s not because the company turned evil. It’s because the incentive structure allowed it. A company that stays great over the long run isn’t one that stays virtuous forever. It’s one whose structure permanently preserves the user’s power to say no.