Course contentsShow
← Data Science
Lesson 230 of 2,1457. Sampling and EstimationFree lesson

Why We Sample Instead of Census

Practical reasons for sampling: cost, time, destructive testing, and when a census is feasible.

Why We Sample Instead of Census

You'll learn the practical reasons data scientists choose sampling over surveying an entire population.

The Core Idea

A census means collecting data from every single member of a population. A sample means collecting data from just a subset. While a census sounds ideal—after all, you'd have complete information—most of the time it's simply not practical. Sampling isn't a compromise; it's often the only realistic option.

Why We Sample: The Practical Reasons

Cost

Surveying millions of people or testing thousands of products costs enormous amounts of money. Sampling a few hundred or thousand units costs a fraction of that while still providing reliable insights.

Time

A census can take months or years to complete. By the time you finish, the data may be outdated. A well-designed sample delivers answers in days or weeks.

Destructive Testing

Imagine testing the lifespan of every lightbulb you manufacture—you'd have none left to sell! When testing destroys the item (crash tests, food quality checks, battery life), sampling is your only option.

Infeasibility

Some populations are infinite or constantly changing (all future customers, water molecules in an ocean). You literally cannot count them all.

When Is a Census Feasible?

Censuses work when populations are small, accessible, and valuable enough to justify the effort. Think: inspecting every part in a critical spacecraft component, or surveying all 50 employees in a small company.

Key Takeaway: Sampling isn't about cutting corners—it's about getting reliable answers efficiently when measuring everything would be too expensive, too slow, destructive, or simply impossible.