What is the law of large numbers?
The more data you gather, the closer your measured result gets to the true value.
Try it
After 50 visitors the observed rate is 8.0%; the true rate is 10%. Keep sliding right to close the gap.
Move the slider to add more visitors and watch the conversion rate settle. The same holds for any percentage you track, such as bounce rate or click-through rate: a small number of visitors tells you very little, while a large number reveals the true rate.
Small sample
The conversion rate you measure in a small sample is unreliable. If the true conversion rate is 10%, one sample of 50 visitors may give a measured rate of 4%, while a second sample of 50 visitors gives a measured rate of 14%.
The figure varies from sample to sample because a small group is easily skewed by a handful of individual visitors, even though the true conversion rate has not moved.
Large sample
The conversion rate you measure in a large sample is dependable. If the true conversion rate is 10%, a sample of several thousand visitors will give a measured rate close to 10%, and it barely changes as more visitors are added.
A large group cannot be skewed by a few individual visitors, so the measured rate stays close to the true conversion rate.