I jumped straight into breaking down the geography before thinking about what actually drives cost.
Break the problem into supply (area to cover) and cost drivers (equipment, labor, processing). Estimate the number of miles of roads in Manhattan, then calculate the number of camera-equipped vehicles and days needed, and multiply by daily operational costs. Finally, add data processing and storage costs to get a total.
Pro tip: State your assumptions clearly and round numbers to make calculations easy; interviewers care more about your structured thinking than the exact figure. Also, mention that Google already has Street View in Manhattan, so this is a hypothetical exercise—showing awareness of existing data.
Confirm that 'all of Manhattan' means every public street, and assume we need to cover both sides of the street. State that we'll ignore one-time setup costs like software development.
Manhattan is about 22 square miles with a dense grid. Estimate total street miles by assuming a grid of avenues and streets, or use population density and average road miles per capita. A reasonable estimate is ~500-600 miles of roads.
Calculate how many vehicles and days are needed. Assume a car can cover 20-30 miles per day (due to slow speeds, traffic, and stops). With 500 miles, that's about 20-25 vehicle-days.
Include vehicle lease/operation, camera equipment, driver salary, and fuel. Assume $500-$1000 per vehicle-day. Multiply by 20-25 days to get $10k-$25k for collection.
Estimate data processing (stitching images, blurring faces) and storage. Assume $0.10 per image and 10 images per mile, so 500 miles = 5,000 images = $500. Storage is negligible. Total cost is dominated by collection.
AI-generated suggestions, not part of the candidate's original notes. May be inaccurate — verify before relying on them.