Frameworks/Estimation
← PreviousBehavioral Interview FrameworkNext →Metrics Framework

Estimation Framework

Fermi estimation techniques for sizing markets and making quantitative estimates with confidence.

When to Use This Framework

Use this framework for: market sizing, back-of-the-envelope calculations, resource estimation, and any question asking you to quantify something without given data.

The 4-Step Estimation Process

Step 1: Clarify the Question

Make sure you understand exactly what you are estimating. Confirm units, scope, and geography.

Ask: "Are we estimating global or US-only? Are we counting daily, monthly, or annual figures? Is this revenue, users, or volume?"

Do not assume. A quick clarifying question prevents a two-minute calculation in the wrong direction.

Step 2: Structure Your Approach Using MECE

Before calculating, break the problem into components that are Mutually Exclusive and Collectively Exhaustive (MECE). This means your pieces do not overlap and together they cover the whole.

Say your structure out loud before calculating. "I'm going to break this into X and Y, then add them together." This signals structured thinking even before you have a number.

Top-down approach: Start from total population and narrow down by segments.

"There are 330M Americans. 60% are adults with smartphones. Of those, 30% use ride-sharing..."

Bottom-up approach: Start from a unit and scale up.

"One Uber driver completes ~20 rides per day. There are ~150,000 active drivers in NYC..."

Use whichever feels more natural. Cross-checking using both approaches and comparing the results is a strong move.

MECE structures for common question types:

Geography: US + international, or by city tier
User segment: by age group, income, behavior
Time: daily × days per year, weekday vs. weekend
Supply side: number of providers × capacity per provider

Step 3: Estimate Each Component

Break the problem into pieces you can estimate independently. Use round numbers — precision is not the point.

Useful reference numbers to memorize:

US population: ~330M
World population: ~8B
US households: ~130M
US smartphone users: ~270M
US internet users: ~310M
Average US household income: ~$75,000
NYC population: ~8M
Average American life expectancy: ~78 years
US adults who drive: ~230M
Average American watches ~4 hours of TV per day

When you do not know a number, make a transparent assumption. "I'll assume 20% adoption — I don't have data on this, but that feels reasonable for an early-stage product. I'd validate this assumption first in a real scenario."

Step 4: Calculate, Then Sanity Check

Multiply your components together. Show your work out loud.

Sanity check: Is the answer plausible? Compare to what you know or to an anchor from a different direction.

"That gives me 2 million rides per day in NYC. Uber reported 5M global trips per day in early years, so 2M for NYC alone seems high — let me revisit my driver count."

Give a range, not just a point estimate. "My estimate is 1-2 million rides per day, with 1.5M as my central estimate." Ranges signal appropriate humility about uncertainty without abandoning rigor.

Catching and correcting errors in real-time is a positive signal to interviewers.

Common Estimation Structures

Market size: Population × adoption rate × average spend

Revenue: Users × conversion rate × average order value × frequency

Capacity: Units × throughput per unit × utilization

Total time spent: Users × sessions per day × average session length

Common Mistakes to Avoid

  • Not clarifying the question before starting
  • Skipping the structural breakdown and jumping straight to numbers
  • Violating MECE — double-counting or leaving gaps in your structure
  • Getting stuck on precision instead of moving forward
  • Skipping the sanity check
  • Giving only a point estimate with no acknowledgment of uncertainty
  • Saying "I don't know" without attempting an estimate
← PreviousBehavioral Interview FrameworkNext →Metrics Framework