Fermi Estimation

Description

Decomposes an unknown quantity into estimable factors, bounding each, to reach an order-of-magnitude answer and identify which assumptions matter most.

Fermi famously estimated the yield of the Trinity test by dropping paper scraps. Fermi estimation breaks a question ('how many users will hit this error each month?') into factors (users per month x proportion using feature x error probability per use x uses per user) and estimates each with plausible ranges. Errors in independent factors tend to partially cancel, giving surprisingly good order-of-magnitude answers. Extending with ranges and Monte Carlo simulation produces distributions and shows which factor dominates uncertainty, guiding research priorities. It disciplines UX sizing claims and opportunity estimates.

Attributes

Loss type
Latent
Stage
Strategy
Effort
Hours

Steps

  1. State the quantity precisely, with units and time frame.
  2. Decompose it into multiplicative or additive factors that are easier to estimate.
  3. Estimate each factor as a range (e.g. 90% interval) from data, research or reasoning.
  4. Combine point estimates for a quick answer, then simulate ranges with Monte Carlo for a distribution.
  5. Identify the factors contributing most to uncertainty (sensitivity analysis).
  6. Measure or research the dominant factors if the decision depends on them.

Sources

  • Weinstein & Adam, Guesstimation: Solving the World's Problems on the Back of a Cocktail Napkin (2008)
  • Hubbard, How to Measure Anything (3rd ed., 2014)

Also known as

  • Fermi problems
  • Order-of-magnitude estimation
  • Back-of-the-envelope estimation

Categories

  • Information Theory, Physics & Mathematics