Anti-Goodhart Entrepreneurial Strategies

Why a founder is exposed to Goodhart three times more than an employee

A founder is exposed to Goodhart on three simultaneous fronts:

1. Investors: they measure your company with standardized KPIs (MRR growth, NRR, magic number, burn multiple) — you are incentivized to optimize them, sometimes against the business itself.
2. Team: you set KPIs for your team, who optimize some of them against the real goal.
3. Yourself: you watch personal dashboards (DAU, MRR, NPS) that shape your decisions, sometimes by breaking the global system.

A lucid founder spends 20% of their time asking themselves which of their dashboards is currently lying. Here is the method.

Step 1: audit your current metric stack

List your 5 to 10 main indicators. For each, answer:

Score 4+ "yes" out of 6 = the metric is probably already infected.

Step 2: build your anti-Goodhart matrix

For each critical metric in your business, build the anti-Goodhart triplet:

Steering becomes rigorous when all three columns point in the same direction. If the offensive column rises but defensive or delayed stagnates / drops → Goodhart in progress.

Step 3: the rule of 3 openings

Borrowed from Andy Grove (Intel) and formalized in modern agile management. Each quarter, open three questions on your KPIs:

1. Which metric is being gamed and nobody is talking about it? Ask 3 ICs anonymously.
2. Which metric has gone up without real value going up? Cross-reference with an external signal (review, NPS, qualitative expert).
3. Which metric will we stop tracking this quarter? To fight dashboard inflation and deep-optimization pressure.

Step 4: the "disposable metrics" policy

A policy adopted by Stripe, Figma and Linear since 2022: rotate KPIs every 2-3 quarters.

Mechanics:

• Q1: main KPI = D7 activation
• Q2: main KPI = M1 retention
• Q3: main KPI = product NPS
• Q4: main KPI = M3 cohort LTV

Why it works:

• Deep optimization takes 6-9 months to install
• Changing the KPI every 2-3 quarters kills gaming before it dominates
• Goodhart risk is bounded to a short horizon

Drawback: the team can lose focus. Use mostly in growth/product, less in sales (where commission-plan stability is critical).

Step 5: align bonuses with long loops

Larkin's rule (Strategic Management Journal, 2014):

Any metric with a feedback loop < 6 months is massively gamed. Any metric with a loop > 12 months is too distant to motivate.

The practical compromise:

• Monthly/quarterly bonus (short term): 30% of variable max — on simple, low-Goodhart-prone metrics (capped call volume, processed tickets).
• Annual bonus (mid term): 50% of variable — on delayed metrics (NRR at 6 months, cohort retention).
• Multi-year bonus (long term): 20% — equity, long vesting, tied to long-term business health.

This structure inverts incentives: short-term gaming pays little; long-term quality pays a lot.

Step 6: install a "Goodhart referee"

An emerging organizational innovation (Asana, Notion, Vercel post-2023): name a Goodhart referee in the RevOps or DataOps team. Their mission, owned:

• Quarterly audit of KPI distributions (anomalies, spikes, drift)
• Read 10 calls / 10 tickets / 10 deals blind per month
• Present to the CEO a quarterly Goodhart Risk Report
• Have the authority to suggest removing or rotating a KPI

This role typically represents 0.3 to 0.5 FTE. Documented ROI: prevention of ~1 internal scandal every 2 years, savings of 2-5% of revenue in avoided gaming.

Practical case: Series A B2B SaaS

Context: HR SaaS, €4M ARR, 30 people, sales team of 6 AEs.

Initial KPI (Q1 2025): signed MRR, monthly bonus.

Goodhart symptoms observed in Q3:

• 90-day churn rose from 4% to 12%
• Onboarding NPS dropped from 47 to 28
• Average discount went from 8% to 21%
• Pipeline "velocity" halved (AEs forced slow deals to close)

Commission plan redesign (Q4):

Result 12 months later:

• 90-day churn: back down to 5%
• Onboarding NPS: 53
• Average discount: 11%
• MRR: -8% in Q1, +18% in Q4

The short-term slowdown was offset by recovered pipeline quality. The system is now resistant to Goodhart.

The lucid founder's golden rule

Any number you steer on will eventually be gamed. The question is not "will it happen", but "how long until it does", and "with which defensive metric will I detect it".

Five habits to adopt:

1. Read 5 customer conversations per week (zoom calls, intercom, support) — qualitative catches what KPIs hide.
2. Have 1 offensive + 1 defensive + 1 delayed KPI on each critical axis.
3. Rotate main KPIs every 2-3 quarters on growth/product functions.
4. Align 70% of annual bonus on loops > 6 months.
5. Each month, ask 1 IC anonymously: "which metric is being gamed?".

The VC trap: Goodhart applied to your company by your investors

A final warning. Your investors put you under Goodhart pressure: they track a standardized dashboard (MRR growth M/M, magic number, burn multiple), and reward or penalize accordingly.

Three classic traps:

1. Push MRR at any cost before a round → aggressive discounts, delayed churn that surfaces post-Series-B.
2. Optimize burn multiple by cutting marketing → revenue loop that collapses 2 quarters later.
3. Hire to show team growth → over-staffed team you'll have to lay off in 18 months.

The countermeasure: share your anti-Goodhart matrix with your board starting from the board deck. Show the offensive KPI, the defensive, and the delayed. Serious investors appreciate the rigor. Short-termist investors expose themselves.

Operational synthesis

You now know how to:

• Detect Goodhart variants (Regressional, Extremal, Causal, Adversarial)
• Build offensive / defensive / delayed KPI matrices
• Align bonuses with long loops (>6 months)
• Rotate KPIs to prevent deep optimization
• Audit your own metric stack quarterly

The final quiz that follows validates this complete understanding.