Design Incentive Systems

---
name: design-incentive-systems
description: Use when designing governance, organizational processes, or operational controls — to audit whether each desired outcome requires individual virtue to produce, and replace virtue-reliance with structural mechanisms that make good behavior the rational self-interested choice
source: '荀子 (Xunzi) 性恶篇 (~3rd century BC) — 人之性恶，其善者伪也 (Human nature is bad; goodness is constructed through institution); Jensen & Meckling "Theory of the Firm" (1976) — principal-agent problem, alignment through incentive structure; Thaler & Sunstein "Nudge" (2008) — choice architecture; Munger "Poor Charlie''s Almanack" — Show me the incentive and I''ll show you the outcome'
tags: [governance, organizational-design, incentives, systems-thinking, institutional-design, behavioral-economics]
related: [apply-goodharts-law, apply-earnings-myopia-defense]
verified: true
---

# Design Incentive Systems

Design governance and organizational systems with the explicit assumption that actors follow incentives rather than virtue — for each desired outcome, identify whether the current structure makes it the rational self-interested choice, and replace virtue-reliance with structural mechanisms.

## Why This Is Best Practice

Xunzi (荀子, ~3rd century BC) made the most controversial argument in classical Chinese philosophy: "人之性恶，其善者伪也" — human nature is inherently bad; its goodness is constructed through artifice and institution (伪/wéi). This was a direct refutation of Mencius, who held that human nature is inherently good and governance merely needed to cultivate it. Xunzi's conclusion was institutional: since goodness is not inherent, governance must not rely on it. Ritual (礼) and law (法) must be designed to produce good outcomes through structural constraint, not through the virtue of the individuals within the system. Good institutions produce good outcomes regardless of the character of the individuals they govern.

**Jensen & Meckling "Theory of the Firm" (1976):** The most-cited paper in the history of economics formalizes the same insight: agents (employees, managers, contractors) act in their own self-interest; principals (shareholders, boards, employers) must align incentives structurally rather than relying on loyalty or virtue. This foundational finding drives VC term sheets, CEO compensation design, option vesting schedules, board independence requirements, and audit structures across every publicly traded company globally.

**Charlie Munger:** "Show me the incentive and I'll show you the outcome." This principle is the explicit operating framework at Berkshire Hathaway. Munger attributes most organizational failures not to individual character failures but to incentive structures that made bad behavior rational. He applies this principle to every organizational design decision, investment evaluation, and regulatory assessment.

**Thaler & Sunstein "Nudge" (2008):** Choice architecture — designing the environment so that the path of least resistance leads to the desired choice — is the modern behavioral economics formulation of the same principle. Used in pension enrollment (default opt-in), organ donation policy (opt-out default), and product design globally. Adopted by the UK Behavioral Insights Team (Nudge Unit), the US Office of Information and Regulatory Affairs, and behavioral design teams at Google, Facebook, and major consumer companies.

**Amazon "Mechanisms over Heroism":** Jeff Bezos explicitly stated that Amazon builds mechanisms — structural processes that produce desired outcomes regardless of individual heroism — rather than relying on exceptional people to do exceptional things. Examples: the PR/FAQ process forces customer-value thinking before engineering investment; the two-pizza team rule forces local ownership and accountability; the S-team narrative format forces clarity before decision. Amazon's operational excellence is explicitly attributed to mechanisms, not to the character of its employees.

**Sarbanes-Oxley (SOX) separation of duties (2002):** The person who authorizes payments cannot be the person who processes them; the person who prepares financial statements cannot audit them. These structural separations make fraud structurally harder regardless of character — required for all US public companies. Post-Enron, this became the standard across global financial reporting.

**Why best:** Virtue-reliance has two structural failure modes: (1) it breaks when individuals change — key person departure, stress, temptation; (2) it is invisible until it fails — systems relying on virtue cannot be audited, monitored, or improved. Structural incentive systems can be audited, monitored, and iterated. They generalize across individuals rather than being person-specific. And they produce consistent outcomes under stress, when virtue is most likely to fail.

**Distinct from `apply-transparent-rules`:** apply-transparent-rules publishes rules explicitly before situations arise so enforcement is predictable and discretion is minimized. Rules deter bad behavior by raising costs; they still require actors to choose compliance over the penalty. design-incentive-systems goes further: it makes good behavior the rational choice even without the rule — the incentive structure itself makes violation irrational. Published rules are a necessary but insufficient condition for good governance; incentive alignment makes rules less necessary by making compliance self-interested.

**Distinct from `apply-institutional-integrity`:** apply-institutional-integrity addresses the moment when an existing rule is violated by someone with high status or personal relationship — requiring the enforcer to act against self-interest to preserve the institution. This skill requires virtue from the enforcer; it is a remedy for virtue-reliance failure. design-incentive-systems is upstream: it prevents that failure mode by redesigning the structure before the test arrives.

**Adopted by:** Every publicly traded company globally (Jensen & Meckling principal-agent framework drives VC term sheets, CEO compensation design, option vesting, board independence requirements); Amazon ("Mechanisms over Heroism" — PR/FAQ, two-pizza teams, S-team narrative format); UK Behavioral Insights Team (Nudge Unit), US Office of Information and Regulatory Affairs, and behavioral design teams at Google and Facebook; Sarbanes-Oxley separation-of-duties requirements (all US public companies post-2002).

**Impact:** Jensen & Meckling's "Theory of the Firm" (1976) is the most-cited paper in the history of economics; SOX separation-of-duties requirements post-Enron became global financial reporting standard; Thaler & Sunstein's opt-out pension enrollment default increased retirement savings participation rates dramatically across adopting organizations.

## Steps

1. **Enumerate the desired outcomes the system is supposed to produce.** List each outcome explicitly: "accurate financial reporting," "customer-first product decisions," "on-time deliveries," "safe equipment maintenance." Do not list aspirations ("high integrity"); list behaviors with observable outputs.

2. **For each outcome, audit the current structure: is it incentive-aligned or virtue-reliant?** Ask: "If a self-interested actor with no loyalty to the organization were placed in this role, would they produce this outcome?" If yes — the outcome is incentive-aligned. If no — the outcome is virtue-reliant. Both categories should exist; the question is which ones are load-bearing.

3. **Identify the highest-stakes virtue dependencies.** Not all virtue-reliance is equally risky. Prioritize redesign for outcomes where: (a) the cost of failure is high (financial, reputational, safety, legal); (b) the person holding the virtue dependency changes frequently; (c) the temptation to deviate is high (personal gain, social pressure, time pressure).

4. **Design structural mechanisms for each high-stakes virtue dependency.** The mechanism replaces the virtue requirement with a structural one. Common mechanism types:
   - **Separation of duties:** The person who makes the decision cannot be the person who benefits from it or audits it (SOX, dual-control nuclear procedures).
   - **Visibility/transparency:** Outcomes are observable by multiple parties, removing the ability to hide failure (public dashboards, third-party audits).
   - **Incentive alignment:** Compensation, evaluation, or promotion depends directly on the desired outcome (equity vesting, performance-linked pay, Bezos's "two-pizza" local accountability).
   - **Default design:** Make the desired behavior the path of least resistance (opt-out defaults, required fields, automated reminders, pre-commitment devices).
   - **Pre-commitment:** Require commitment to the outcome before the temptation arises (PR/FAQ before engineering; annual budget locked before Q4 pressure).

5. **Test each mechanism against a simulated adverse actor.** For each mechanism, ask: "If someone in this role specifically wanted to produce the bad outcome, how would they circumvent this mechanism?" A mechanism that can be easily circumvented by a motivated adverse actor is not a structural control — it is a speed bump. Redesign until circumvention requires either collusion (multiple actors) or observable violation.

6. **Preserve virtue-reliance where structural mechanisms are disproportionate.** Not every outcome requires structural incentive alignment. Low-stakes, low-frequency, or highly individualized situations may be better governed by hiring for character and culture than by structural mechanisms. The skill is in identifying where structural mechanisms pay for themselves, not in eliminating all virtue-reliance.

7. **Audit the system as roles and people change.** Incentive systems can drift when reporting structures change, when compensation structures are modified, or when new roles are created. Schedule periodic audits: "For each load-bearing outcome, is the incentive alignment still present under the current structure?"

## Rules

- "We hired good people" is not a governance mechanism. Good hiring reduces the base rate of character failures; it does not eliminate them. Structural mechanisms must not require individual virtue to function.
- Mechanisms that are theoretically robust but practically ignored provide false safety. A separation-of-duties requirement that is routinely bypassed under time pressure is worse than no requirement — it creates the appearance of control without the reality.
- Incentive systems can produce perverse outcomes if poorly designed. Measuring the wrong output and linking compensation to it (Goodhart's Law: "When a measure becomes a target, it ceases to be a good measure") is worse than no measurement. Specify outcomes, not metrics, and monitor for metric gaming.
- Overcorrecting for virtue-reliance destroys culture. Organizations with excessive structural controls signal distrust and suppress the discretionary effort that rules cannot specify. Balance: structural controls for load-bearing outcomes; culture and hiring for everything else.

## Examples

**Financial controls:** A startup scales from 10 to 100 employees. Finance processes previously relied on the CFO's personal oversight of all transactions. As scale grows, a single CFO cannot monitor all transactions — the system becomes virtue-dependent at scale. Design: require two signatures above $10K; separate payment authorization (CFO) from payment processing (controller); third-party audit quarterly. Now each control works without requiring a specific individual's virtue.

**Product organization:** A product team is supposed to prioritize customer value, but product managers are evaluated on features shipped. This incentive structure makes feature velocity the rational career choice, regardless of customer value. Redesign: evaluate PMs on customer retention and NPS impact, not feature count. Now customer-value decisions are incentive-aligned.

**Sales commission:** Salespeople are commissioned on revenue at close. This makes closing deals the rational choice, even if the customer is a poor fit who will churn and generate support load. Redesign: commission on 12-month retained revenue. Now the salespeople's self-interest is aligned with customer success, not just initial close.

**Engineering safety:** A manufacturing plant relies on operators to self-report equipment safety issues. Self-reporting is virtue-reliant: an operator who fears the production slowdown of a maintenance stop has an incentive not to report. Redesign: independent safety inspections, anonymous reporting with guaranteed non-retaliation, and maintenance downtime not tracked against operator performance metrics. Safety reporting becomes incentive-neutral.

**Board governance:** A company's board relies on the CEO to provide complete and accurate information for strategic decisions. This is virtue-reliant — the CEO selects what the board sees. Design: independent board access to department heads, mandatory rotation of external auditors, compensation committee independence from CEO influence. Information quality now depends on structure, not CEO character.

## Common Mistakes

- **Designing rules without incentive alignment:** Transparent rules (apply-transparent-rules) are necessary but insufficient. A rule against nepotism in hiring does nothing if the hiring manager is evaluated on team performance (and nepotism candidates are perceived as loyal and manageable). The incentive must also align.
- **Treating good hiring as a structural control:** Hiring for character reduces base rates; it does not eliminate character failure under sufficient temptation. The executive who defrauds a company typically had an excellent character profile. Structural controls must operate independently of individual character.
- **Over-specifying mechanisms at low stakes:** Designing separation-of-duties protocols for the coffee budget is disproportionate and signals distrust without reducing meaningful risk. Apply structural design proportionally to the stakes.
- **Not auditing for drift:** Incentive systems degrade as organizations change. A well-designed incentive system in year 1 may be misaligned by year 3 if reporting structures, compensation plans, or business models have changed. Build periodic audits into the system design.
- **Goodhart's Law ignorance:** Linking compensation to a specific metric incentivizes optimizing the metric, not the underlying outcome. Sales revenue (metric) and customer value (outcome) diverge when salespeople can sell bad-fit customers. Specify the outcome; choose the metric carefully; monitor for gaming.