When conducting A/B testing analysis, the test you ran is often referred to as a treatment, which is the variable you changed between the test (treatment) group and the control group. Treatment effects include three metrics that capture the impact of the treatment variable you are testing:
- Unadjusted Impact: this is the raw, unadjusted difference in outcomes between the treated group and the control group. It is typically biased by all sorts of factors that were uncontrolled and unrelated to the outcome being measured. A better measure is the ATT defined below.
- Average Treatment Effect on the Treated (ATT): the ATT is the estimated average change in outcome for the treated group as a result of applying the treatment vs. if you had not. This is generally the most reliable estimate as the methodology adjusts as much as possible for all other factors to make this an apples-to-apples comparison between the treated group and the control group.
- Average Treatment Effect (ATE): this is the average change in your outcome variable (i.e. dependent variable) across the entire population if you applied the treatment to all in your dataset. It is often biased by a variety of factors you did not control or account for (confounding factors).
- Average Treatment Effect on the Control (ATC): the ATC is the estimated average change in outcome for the control group as a result of applying the treatment vs. if you had not.
Note that these effects are estimated using the propensity-matched treated and control groups. Analyzr also provide standard error estimates (SE) and sample sizes (N) for these groups in the left-hand side table