ml

๐ŸŸ  Product Sense: Fermi Estimation

These test structured thinking, not exact numbers.

The framework 1. Clarify scope: Define what you're estimating 2. Break into components: Multiplication chain of estimable pieces 3. Estimate each component: Round numbers are fine 4. Sanity check: Does the result feel reasonable?

Worked example: "How many Uber rides happen in Chicago per day?"

Approach 1 (demand-side): - Chicago metro: ~9.5M people - Maybe 10% use ride-sharing monthly: 950K - Average ride-share user: ~4 rides/month รท 30 days โ‰ˆ 0.13/day - 950K ร— 0.13 โ‰ˆ ~125K rides/day

Approach 2 (supply-side): - Estimate ~30K active drivers in Chicago - Each driver does ~8-10 rides per shift - Maybe 60% are active on a given day: 18K - 18K ร— 9 โ‰ˆ ~160K rides/day

Two approaches giving roughly similar answers โ†’ confidence we're in the right ballpark.

Practice these on your own: - How many text messages are sent in the US per day? - How much revenue does a single Starbucks generate per year? - How many data scientists are there in Chicago?