Sanity Checks for Fermi Estimation
Sanity checks for Fermi estimation interviews, including units, bounds, benchmark comparisons, sensitivity, and communication.
Candidates who produce estimates but struggle to validate them.
Check the units
Every Fermi estimate should end in the requested unit. If the prompt asks for dollars per year, do not accidentally answer users per month.
Use bounds
Create a rough lower and upper bound. The final estimate should sit inside that range and move in the expected direction when assumptions change.
Concrete example
If estimating daily rides in a city, the answer cannot exceed population times plausible trips per person. A result above that bound signals a unit or frequency error.
Test sensitivity
Change the most uncertain input and see how much the final answer moves. If one assumption dominates, discuss it explicitly.
Common mistakes
Candidates often stop after multiplication. A Fermi answer without a sanity check is just arithmetic on assumptions, so always test scale, units, and the dominant driver.
Practice the pattern
Use the LeetQuidity curriculum and calibration to turn this topic into a focused practice plan.