 Randomized control tests don't necessarily tell us how much effect a treatment has. In the test, a sample's outcome depends on the observed variables that we know have treatment effects and the unobserved variables that may or may not have treatment effects. Average treatment effect for a test is measured by taking the mean of the treatment group output and subtracting the control group outputs. And doing some math, this leads to a final form with this error term that can be huge. This can be lowered with larger sample sizes. And this beta term is an unbiased estimate of treatment effect that may not be accurate, but it can become more precise by dealing with confounding variables and also following the assumptions of ignorability and SITFA among others. Just be careful, references down below for more details.