When to write a policy vs. when to have a conversation.

Someone takes a four-hour lunch. A founder writes a lunch policy. Someone misses a deadline. A founder writes a Slack-response-time policy. Someone takes too much vacation. A founder writes a vacation cap. Within 18 months, you've built a company that runs on rules — and the rule-followers stay, while the people you actually wanted to keep quietly leave.

Policy is a powerful, expensive lever. It scales — which is the point, and also the problem. Conversation is cheap, fast, and specific. Most leaders pull the wrong one. Here's the test.

The four-question policy test

Before you write a new rule, ask these four questions in order. If you answer 'no' to any of them, don't write the policy. Have the conversation.

• 1. Has this happened more than three times, with three different people?
• 2. Is the harm real and measurable — not just 'it bothered me'?
• 3. Will the cost of enforcement on everyone be lower than the cost of the behaviour on the few?
• 4. Can you write the rule in one sentence that a stranger would understand the same way you do?

The psychology: norm-setting vs. rule-setting

Cialdini's research on social proof shows that humans calibrate behaviour to what they perceive others doing. A well-handled conversation about one person's behaviour broadcasts a norm to the whole team without anyone needing to read a document. A policy, by contrast, says 'we expect this to keep happening, so we're institutionalising the response.' That's a very different cultural signal — and people receive it loud and clear.

When policy is exactly the right answer

• Safety, legal, regulatory, or fiduciary risk. Always policy. Always written. Always reviewed by a lawyer.
• Decisions that need to be the same across hundreds of cases for fairness (parental leave, comp bands, security access).
• Anything where a missing rule creates structural inequity (e.g. promotion criteria, hiring loops).
• Things you genuinely don't want any individual manager to be able to override (e.g. ban on unpaid trial work).

When conversation is the right answer (and policy is the wrong one)

Real case: the vacation policy that cost a team

A 25-person company had unlimited PTO. One employee took 8 weeks in a year. The founder, alarmed, wrote a 4-week cap. Within a quarter, average PTO across the company dropped from 14 days to 9. Two senior engineers left, citing 'creeping rigidity.' Replacement cost: ~$280K. The original 8-week employee was, in fact, going through cancer treatment they hadn't disclosed. One conversation would have solved everything. One policy broke the company's culture for two years.

Take this home — the policy filter

• Add a 7-day cooling-off rule for yourself: never write a new policy in the week you're upset.
• Maintain a 'conversations had' log. If the same issue shows up 3+ times with 3+ people, escalate to policy review.
• When you do write a policy, write the conversation it replaces too. Share both with the team.
• Sunset every policy after 18 months. Renew or delete. Default to delete.
• Once a year, audit your handbook with a new hire. Ask them to flag any rule that doesn't make obvious sense. You'll be surprised.