A city has millions of households. You will ask a few hundred of them one question — and get remarkably close to the right answer. That is not luck. Draw sample after sample below and watch the estimates cluster around the truth, then find out what makes the whole trick collapse.
Where the estimates land
one sample’s estimatethe truth, 62%your latest sample
Your latest sample, household by household
Press “Draw one sample” to knock on some doors.
How precision grows with sample size
margin of error (95%)bias — the part n cannot fix
Ask a few hundred householdspress the button below
200
1 million
Samples drawn0
Latest estimate—
Off the truth by—
Centre of the pile—
Margin of error (95%)—
Within MoE of truth—
Precision comes from n, not from N. A spoonful tells you about the pot whether the pot holds a litre or a swimming pool — provided you stirred first. Stirring is what randomness does.
Randomness is not a formality. Switch to Convenience sample and slide n up to 2,000. The estimates get tighter and tighter — around the wrong number. A biased method does not become less wrong with more data; it becomes more confidently wrong.