Goals, Incentives, and Accountability

The managerial question: why smart people do the wrong things

Every organization eventually discovers that its people are not stupid, lazy, or indifferent. They are rational. They respond to the goals they are given, the incentives they are paid for, and the accountability systems that determine whether anyone notices when outcomes drift. When those three elements pull in different directions, the result is not random failure. It is predictable, optimized behavior that destroys value.

Consider a pattern that repeats across industries. A SaaS (software as a service) company declares that customer success is strategic. The sales team is compensated on bookings (signed contract value). Customer success is compensated on gross retention (keeping existing revenue). Product is measured on features shipped. Finance celebrates margin expansion. Each function hits its numbers. Churn rises. Implementation backlogs grow. Executives ask why "everyone performed" while the business deteriorated.

The answer is not culture. It is design. Goals tell people what to optimize. Incentives amplify the optimization. Accountability determines whether variance is surfaced early or buried until quarterly earnings. When a manager inherits a broken triad, the fix is not a motivational speech. The fix is to redesign the system so that doing the right thing for the company is also the rational thing for the individual.

This lesson teaches you to design and diagnose that triad. You will learn how OKRs (Objectives and Key Results, a goal-setting framework that separates qualitative direction from measurable outcomes) align effort without becoming task lists. You will learn why Goodhart's law ("when a measure becomes a target, it ceases to be a good measure") makes single-metric incentives dangerous. You will learn how accountability differs from blame, and why named owners with review cadence matter more than posters on the wall. From Unit 3, you already know how to decompose problems and prioritize by impact. This lesson connects that analytical clarity to the human systems that must execute it.

Goals: what the organization is trying to achieve

A goal is a statement of desired outcome that allocates attention. Without explicit goals, people default to local optimization: the metric their boss asks about in the weekly meeting, the project with the loudest stakeholder, the work that is easiest to demonstrate. Goals are not wishes. They are attention budgets expressed as outcomes.

Good goals share four properties. First, they are outcome-oriented, not activity-oriented. "Increase enterprise net revenue retention to 115%" is an outcome. "Launch three features" is an activity that may or may not move retention. Second, they are few enough to force tradeoffs. A team with twelve "top priorities" has no priorities. Third, they are time-bound, so progress can be judged without infinite deferral. Fourth, they connect to a causal chain the team understands: if we hit this key result, why should the business improve?

OKRs formalize this pattern. An Objective is qualitative and directional: it answers "where are we going?" A Key Result (KR) is measurable and time-bound: it answers "how will we know we arrived?" The framework was developed at Intel under Andy Grove and popularized at Google. It spread because it solves a common failure mode: confusing tasks with outcomes.

At Google-scale, OKRs are set quarterly with weekly check-ins. Teams score key results on a 0.0 to 1.0 scale, with 0.7 often treated as success on ambitious targets. The scoring convention matters: OKRs are meant to stretch. If every team scores 1.0 every quarter, the organization is either sandbagging or the goals are tasks disguised as outcomes.

OKRs fail in predictable ways. Teams write key results that are deliverables ("ship migration tool") instead of outcomes ("reduce median deployment time from 45 minutes to 15 minutes"). Executives cascade too many objectives, diluting focus. OKRs are tied directly to bonus formulas, which encourages conservative targets because missing a bonus-linked KR has direct financial pain. The fix is discipline: three to five key results per objective, outcome language, and separation between OKRs (alignment and learning) and compensation (fairness and retention).

Managers should treat goals as hypotheses, not commandments. A KR that stays red for six weeks is information: the strategy may be wrong, the resourcing may be insufficient, or the KR may measure the wrong thing. The weekly check-in exists to update beliefs, not to perform optimism.

Incentives: what behavior gets rewarded

If goals define the destination, incentives define the fuel mix. An incentive is any reward (money, promotion, status, autonomy) tied to measured performance. People respond to incentives with remarkable precision, including when the response harms the company. That is not cynicism. It is economics applied inside the firm.

Incentive design starts with a simple question: what outcomes does this role control? A retail store manager controls staffing schedules, local merchandising, and customer recovery within policy. She does not control national advertising spend or supplier negotiations. Compensating her on company-wide operating margin mixes controllable and uncontrollable factors, which feels unfair and encourages gaming of whatever piece she can influence.

Strong designs balance leading and lagging indicators. Lagging indicators (revenue, profit, retention) tell you whether you won. Leading indicators (pipeline quality, defect rates, onboarding completion) tell you whether you are on track to win. Incentives tied only to lagging indicators encourage end-of-period heroics and destructive shortcuts. Incentives tied only to leading indicators encourage activity without results. Pair them.

Delay matters. Quality often reveals late. A sales team paid on bookings in the quarter may sell deals that churn in month four. Clawbacks (repayment of commission if the customer cancels within a window) and multi-year vesting on variable pay align the time horizon of reward with the time horizon of value. Without delay, you pay for signatures, not relationships.

Team and individual components matter when work is interdependent. Pure individual incentives in a cross-functional product launch encourage finger-pointing. Pure team incentives allow free-riding. A common pattern is 70% team outcome (launch success metric) and 30% individual contribution (peer-rated or manager-rated against role-specific behaviors).

The most famous incentive war story of the last decade is Wells Fargo (2016). Branch employees faced aggressive cross-sell targets: open a certain number of new accounts per customer. The goal was measurable. The incentive was direct (job security, bonuses). The accountability system punished missing targets. Employees responded rationally by opening accounts customers never requested. Millions of unauthorized accounts were created. Regulators fined the bank billions. Leadership blamed "bad apples." Analysts blamed incentive design: when the measure (accounts opened) became the target, it ceased to be a good measure of customer relationship health.

A softer but equally instructive war story comes from enterprise software sales. A company paid account executives on annual contract value (ACV, the yearly value of a signed subscription) with no quality filter. AEs sold multi-year deals to customers outside the ICP (ideal customer profile, the segment the product serves best). Bookings soared. Implementation timelines stretched. Customer success managers inherited accounts they would never have sold. Gross retention fell. Sales blamed success for poor onboarding. Success blamed sales for bad fit. The war was inevitable because incentives optimized conflicting proxies for "growth."

The fix was not a values statement. It was a redesigned metric: quality-adjusted bookings, where ACV counted at full value only if the account met ICP criteria and customer success signed off on implementation capacity before close. AEs still hunted revenue. They also hunted fit.

Goodhart's law and metric pairing

Goodhart's law, named for economist Charles Goodhart, warns that any metric used for control will be gamed. The moment a number determines bonus, promotion, or survival, people optimize the number. Sometimes that optimization aligns with value. Often it does not.

Classic examples appear in every function. Customer support measured on tickets closed per hour produces quick closes and high reopen rates (the same customer returns because the issue was not solved). Schools measured on standardized test scores produce teaching-to-the-test and narrowed curricula. Engineering teams measured on story points completed inflate point estimates and ship fragile code. Marketing measured on leads generated buys low-intent lists that sales rejects.

The managerial response is not to abandon measurement. It is to pair metrics so that gaming one exposes harm on another. Pair quantity with quality. Pair speed with rework rate. Pair bookings with 90-day retention cohort. Pair feature output with defect escape rate.

Dashboards help when they show pairs side by side. Single KPI (key performance indicator, one headline number) culture invites Goodhart behavior. Rotation also helps: emphasize different paired metrics by quarter so no single proxy becomes the permanent target.

When you design an OKR key result, ask: "If my team hit 100% on this KR, what could still go wrong?" That question surfaces the paired metric. If the KR is "reduce average handle time on support calls," the paired metric might be "customer satisfaction on resolved tickets" or "reopen rate within 7 days." If both move in the right direction, you likely have real improvement. If handle time falls while reopens spike, you have gaming.

Accountability: ownership without blame theater

Accountability means clear ownership for outcomes, explicit review of variance, and consequences that follow performance over time. It does not mean public humiliation in meetings or blame shifting when a metric misses.

Diffused responsibility is the default in large organizations. "We all own customer experience" sounds inclusive. In practice it means no one owns it. RAPID (a Bain framework: Recommend, Agree, Perform, Input, Decide) and RACI (Responsible, Accountable, Consulted, Informed) exist to name roles. This lesson focuses on accountability in reviews; Lesson 2 covers decision rights in depth. For now, the key distinction is: accountability requires a named owner who presents the metric, explains variance, and proposes corrective action.

A healthy accountability rhythm looks like this. Each week, the owner of a metric presents: (1) current value vs plan, (2) drivers of variance, (3) actions taken since last review, (4) decisions needed from leadership. Peers provide input. The decision maker commits resources or accepts the miss. The next week, the first agenda item is whether last week's commitments happened. Closed loop accountability is rare and valuable.

Accountability without blame requires separating performance from personhood. Missing a target because the market shifted is different from missing because the owner ignored leading indicators for six weeks. Consequences should follow patterns, not single data points. High-performing cultures still fire people who repeatedly miss commitments without learning. They also protect teams that miss ambitious OKRs while running disciplined experiments.

War story: a consumer fintech set an OKR to "grow active users 40% this quarter." Marketing hit the KR through paid acquisition with low activation. Product's activation KR missed. Neither team "owned" unit economics. Finance discovered customer acquisition cost (CAC, the average spend to acquire one customer) had doubled only at quarter end. Accountability existed on siloed KRs. No one owned the paired outcome: active users at sustainable CAC. The retrospective added a shared KR on "30-day retained actives per dollar of acquisition spend" with a single executive owner who could reallocate budget weekly.

NovaLedger sells accounting automation to mid-market companies. Annual recurring revenue (ARR, the yearly value of active subscriptions) is $48 million. Churn has risen from 8% to 14% over two years. The CEO suspects incentive misalignment.

Part A: Current state (fact pattern)

Q3 results:

Win/loss notes show 38% of new logos were outside ICP (too small, wrong industry). AEs rushed year-end discounting to hit accelerators. CSMs escalated "surprise" churn on accounts they did not sell.

Part B: Diagnose the triad

Goals: Each function had a different implicit goal. Sales goal = sign contracts. Success goal = keep revenue. No shared goal on fit or implementation capacity.

Incentives: AE accelerators kicked in at 110% of quota, rewarding volume over fit. CSM pay did not penalize preventable churn from bad-fit sales. Implementation was measured on throughput, not backlog reduction.

Accountability: Weekly sales standup reviewed pipeline. Success reviewed churn in a separate meeting. No forum paired bookings with 90-day retention cohort.

Goodhart risks: Bookings KR gamed via discounting and non-ICP deals. On-time implementation KR gamed by deferring complex customers (backlog grew).

Part C: Redesign (proposed Q4 system)

Shared OKR (company level):

• Objective: Grow durable recurring revenue in our ICP.
• KR1: Net revenue retention (NRR, revenue from existing customers including expansion minus churn) ≥ 108% in ICP segment.
• KR2: Median implementation start within 3 weeks of signature for ICP deals.
• KR3: Bookings ≥ $3.8M ACV with ≥ 85% ICP mix.

Incentive changes:

Accountability: Single WBR (weekly business review) slot: "New logo health" owned by VP Revenue, presenting cohort retention at 30/60/90 days alongside bookings.

Check: Incentive weights sum to 100% per role ✓. Shared OKR has one owner per KR ✓.

Part D: Managerial read

The board should ask: "What is our cost of growth?" Bookings alone flattered Q3. Pairing bookings with cohort retention and ICP mix reveals whether NovaLedger bought revenue or earned it. The CEO should not announce the fix as "sales behavior problem." It is a system design problem. Lesson 3 will show how the WBR cadence enforces the new accountability loop.

HarborMart, a 120-store grocery chain, rolled out OKRs to district managers. Year one failed: every district scored 0.95 to 1.0 on KRs. Same-store sales were flat. District managers admitted they set KRs they could hit in week two.

Part A: Bad OKR example (District 7, Q2)

Objective: Operate excellent stores.

Activities, not outcomes. No customer or sales linkage.

Part B: Redesigned OKR (District 7, Q3)

Objective: Win trips in our trade areas through reliable in-stock and service.

Corporate capped district OKR scores at 0.7 for bonus eligibility unless corporate NRR-equivalent (EBITDA margin) held. OKRs were decoupled from 100% of variable pay: 60% financial scorecard, 40% OKR learning score.

Part C: Q3 results

Weighted OKR score: 0.76. District missed in-stock but beat on sales and satisfaction. Leadership reallocated one logistics analyst to District 7 for 90 days (resource accountability).

Check: Weights sum to 100% ✓. Scores between 0 and 1 ✓.

Part D: Managerial read

Decoupling OKRs partially from bonus reduced sandbagging while keeping financial discipline. The district manager learned that in-stock was the binding constraint on the sales KR. Without paired customer and sales metrics, in-stock alone could have been gamed by over-ordering perishables. The operator lesson: OKRs work when they force tradeoffs visible to leadership, not when they become compliance checklists.

Meridian Health operates 40 urgent-care clinics. The CEO wants to reduce wait times without harming clinical quality. Current state:

• Clinic managers bonused on patient throughput (patients seen per hour)
• Quality measured by quarterly audits (lagging, not in bonus)
• Average wait time: 47 minutes (target 30)
• Patient satisfaction: 3.6 / 5 (target 4.0)
• Audit score: 88% (target 92%)

Tasks:

1. Write one Objective and three Key Results for Q1 at the clinic network level. Use outcome language.
2. Identify one Goodhart risk per KR and propose a paired metric.
3. Redesign the clinic manager incentive in one table (metric, weight, rationale).
4. Who should own the weekly accountability review, and what four items must they present?

Solution

1. Sample OKR

Objective: Deliver timely, high-quality urgent care patients trust.

2. Goodhart risks and pairs

3. Incentive redesign

Removing throughput from bonus directly addresses the war story pattern: throughput alone rewarded speed over care.

4. Accountability

Owner: Regional medical director (single D for clinical-operational tradeoffs).

Weekly presentation: (1) wait time vs plan by clinic, (2) satisfaction and complaint trend, (3) audit prep status and last week's corrective actions, (4) staffing or supply decisions needed before next week.

Full paragraphs matter here: the regional director must be able to reallocate nurses across clinics when wait time KR is red without waiting for monthly finance review. That speed is why accountability cadence matches the KR time horizon.

A B2B marketplace connects landlords and corporate tenants. Product wants OKRs on gross merchandise value (GMV, total transaction value on the platform). Trust and safety wants OKRs on fraudulent listing takedown time. Sales wants OKRs on new landlord accounts. Q4 results: GMV up 30%, fraud incidents up 50%, enterprise tenant NPS (net promoter score, willingness to recommend) down 12 points.

Tasks:

1. Explain in prose why the three teams optimized conflicting outcomes.
2. Propose one shared Objective and two KRs that force cooperation.
3. Should OKR achievement pay 100% of variable compensation? Explain why or why not.

Solution

1. Conflict explanation

Each team faced a metric that rewarded local optimization. Product's GMV KR encouraged volume without verifying listing quality. Trust and safety optimized takedown speed on fraud already live, not prevention at listing creation. Sales optimized new landlord signups without filtering bad actors. No KR measured tenant trust or fraud-adjusted GMV, so cooperation was optional. Incentives made non-cooperation rational.

2. Shared OKR sample

Objective: Grow trusted transaction volume enterprise tenants rely on.

• KR1: Fraud-adjusted GMV (GMV minus chargebacks and verified fraud) ≥ $22M, up 15% QoQ.
• KR2: Enterprise tenant NPS ≥ 45 (from 38).

Product, trust, and sales share scoring on both KRs at 30% of their OKR weight; remaining 70% stays function-specific but cannot exceed 0.8 OKR score if shared KRs miss.

3. Compensation link

OKRs should not pay 100% of variable compensation. Full linkage encourages sandbagging and hides diagnostic misses (a team might miss an ambitious KR while improving the business). A typical split is 50 to 70% on financial or role scorecard and 30 to 50% on OKR learning, with caps when ethics or compliance KRs fail. Here, fraud incidents rising 50% should trigger a compliance gate: no variable pay above target if fraud rate exceeds threshold, regardless of GMV.

Check: Shared KRs require joint failure modes to surface in WBR ✓.

Key takeaways

• Goals, incentives, and accountability must be designed together; fixing one in isolation rarely works.
• OKRs align on outcomes when key results are measurable, few, and separated from sandbagging incentives.
• Goodhart's law makes single-metric pay dangerous; pair quantity with quality and delay payouts when quality reveals late.
• Accountability requires named owners, weekly variance review, and closed-loop follow-through, not blame theater.

1. Pick a team you know. Write one objective and three key results. For each KR, name the Goodhart risk and a paired metric.
2. Audit one incentive plan: what behavior does it rationally produce if everyone hits target?
3. Continue to Lesson 2: Decision Rights and Organizational Coordination.