Founder School
โ† All modules
Mindset25 min

How to Operate

by Keith Rabois (CS183B) ยท read the original

Keith Rabois ran operations at PayPal, LinkedIn, and Square. His Stanford CS183B lecture compresses that into a system: think of the company as a machine you build, manage like an editor rather than a writer, delegate by task-relevant maturity, hire barrels not just ammunition, and instrument everything with paired metrics.

  1. 1

    Your Company Is the Machine

    Stop thinking of the app as your product. The company is the product, and your output as a leader is measured by what that machine produces without you touching it.

  2. 2

    You're an Editor, Not a Writer

    Manage with a red pen: relentlessly simplify what the team produces. If the volume of editing you do isn't shrinking over time, something is wrong with your hiring or your teaching.

  3. 3

    Ask Questions, Allocate Resources, Keep One Voice

    Clarify with consistent questions, and remember that moving people is the truest form of strategy. The goal is a company that speaks in one voice without you writing every word.

  4. 4

    Delegation and Task-Relevant Maturity

    Delegate by task-relevant maturity, not by how smart someone is. And when you disagree, weigh your conviction against the consequence before overriding.

  5. 5

    Barrels vs. Ammunition

    Company output scales with the number of barrels, people who can carry an idea from conception to shipping. Find them by expanding responsibility until it stops working.

  6. 6

    Dashboards and Paired Metrics

    Build a dashboard everyone actually checks, and pair every metric with its natural counterweight so optimizing one number can't quietly wreck another.

  7. 7

    Transparency Compounds Judgment

    Share context aggressively, board decks included. People can only make the decisions you'd make if they can see what you see.

  8. 8

    Broken Windows and Interesting Anomalies

    Small visible standards set the tone for everything (broken windows), and strange data points deserve investigation, not dismissal. Anomalies are strategy tips from reality.