TB: How Big Things Get Done (Flyvbjerg & Gardner)

Core Thesis

Large projects fail predictably and for understandable reasons — primarily bad incentives and optimism bias. Reference class forecasting (using base rates from similar past projects) is a simple, proven, and underused corrective.

Key Takeaways

  • Think slow, act fast: invest heavily in planning before committing; execute quickly once committed.
  • Reference class forecasting: instead of using your specific project's estimates, look at the distribution of outcomes for similar past projects. Better on biases, better on unknown unknowns, simple and proven.
  • Most large project failures can be explained by bad incentives — skin in the game would improve results significantly.
  • Private sector projects generally have better outcomes than public sector ones.
  • Do the estimates come with intervals? If not, they are not honest estimates.

Mental Models

  • Incentives Matter — bad incentives explain most large project failures
  • Inversion — think about what makes projects fail, not just what makes them succeed
  • Absence Blindness — optimism bias is inability to see the reference class of failures

Source note