Charlie Munger, Warren Buffett's longtime partner at Berkshire Hathaway, kept a one-line rule for predicting how any system will actually behave: "show me the incentive, and I'll show you the outcome." Not the mission statement, not the values deck, the incentive.
His favorite proof of it was Xerox. Early on, Joe Wilson noticed Xerox's better, newer copier was selling worse than its inferior older model. The reason wasn't the product, it was the commission structure: salesmen earned more selling the older machine, so that's what they sold (Kyle James, "Show Me the Incentive"). The sales team wasn't sabotaging the company. They were following the incentive exactly as designed.
The pattern shows up everywhere in marketing
Marketing runs on incentive structures nearly as often as sales does, and most of them were never stress-tested for what they'd actually reward.
Take a commission plan built around total revenue with reps holding price discretion: the predictable result is reps racing to the lowest allowed price to close volume, revenue spikes, and profit tanks, exactly the trap Munger's framework predicts (Kyle James). Nobody designed that outcome on purpose. The comp plan designed it for them.
Three marketing-specific versions of the same trap:
- Agency fee structures billed on hours worked incentivize slower work, not better work, because the agency's revenue depends on the clock, not the outcome.
- Affiliate payouts based purely on click-through or last-touch attribution invite exactly the fraud you'd expect: PayPal's Honey extension was found in December 2024 to be quietly swapping out legitimate creators' affiliate cookies for its own at checkout, redirecting commissions that weren't earned (Affiverse). Industry-wide, an estimated 17% of affiliate traffic is fraudulent, costing businesses roughly $3.4 billion a year.
- Growth team OKRs tied to top-of-funnel signups reward volume, so the team ships every dark-pattern trick that inflates signups, even when it tanks activation downstream.
None of these people are villains. They're rational actors reading the incentive correctly.
Munger's point wasn't cynicism about people, it was realism about systems. Assume competence and self-interest, then ask what the incentive actually rewards.
When the same trap plays out at scale
The Wells Fargo cross-selling scandal is the textbook version of this, and it wasn't a marketing team, but the mechanics are identical to any growth incentive gone wrong. Wells Fargo ran an internal "eight is great" push, telling employees to sell eight financial products per customer, with quotas that carried over and compounded if a rep fell short one day (Don on Selling).
Employees facing an ever-tightening quota with no room to miss did exactly what the incentive rewarded: they opened millions of accounts customers never asked for, just to hit the number. The bank eventually paid $185 million in penalties, with a $3 billion settlement following in 2020 once the pattern was fully unwound (Fiscal Report). The root cause diagnosis from the aftermath applies directly to marketing: the plan rewarded hitting a metric with no regard for how it got hit.
Swap "accounts opened" for "signups captured," "meetings booked," or "leads generated" and the same failure mode is sitting in plenty of ordinary marketing dashboards, just at a smaller, quieter scale.
A mid-size SaaS company paid affiliates a flat $40 per free-trial signup, no conversion requirement. Signups tripled in two months, budget followed, but trial-to-paid conversion fell from 18% to 6% because affiliates were driving traffic from unrelated giveaway sites. The company switched to $15 per signup plus $60 per paid conversion at 30 days. Signup volume dropped by half, but paid customer volume rose 22% on a smaller budget, because the incentive finally matched the outcome the company actually wanted.
Reading a comp plan like Munger would
Before you launch any incentive structure, whether it's an affiliate program, an agency contract, or a bonus tied to a KPI, run it through one question: if someone optimized ruthlessly for this exact metric and nothing else, what would they do?
Try it on a few common structures:
- Pay affiliates on last-click conversion only β reward cookie-stuffing and coupon-site poaching over genuine influence.
- Pay agencies for hours billed β reward slow, padded work over efficient work.
- Pay SDRs on meetings booked, not meetings held β reward booking junk meetings that no-show.
The fix isn't "trust people to do the right thing." The fix is closing the loophole before it opens.
Designing the incentive, not patching the outcome
Once you can see the loophole, redesign around it instead of policing it after the fact.
- Pay for the outcome you actually want, not the easiest-to-measure proxy for it. Affiliate commissions weighted toward verified new-customer revenue, not raw clicks.
- Cap the downside of gaming it. A revenue-based sales bonus with a minimum-margin floor closes the "sell at any price" loophole Xerox and the discount-racing example both hit.
- Audit quarterly, not once at launch. Incentive structures decay as people find the gaps; what worked in January can be exploited by June.
- Involve someone outside the team that benefits from the incentive. The people closest to a comp plan are the least likely to spot its loopholes, precisely because they're the ones who'd exploit them first.
Every incentive structure is a message about what actually matters, whatever your mission statement says. Read your own comp plans the way Munger read Xerox's, and you'll usually find the outcome before it happens.
The Wells Fargo and Xerox stories both share the same root cause: an incentive that was easy to measure standing in for an outcome that was hard to measure. Revenue is easy to track, "did we sell the right product to the right customer" is not; clicks are easy to count, "did this affiliate actually influence the purchase" is not. Whenever you catch a comp plan optimizing for the easy number, that's usually the exact spot where the real damage is quietly accumulating.