We develop a framework that allows researchers to describe a wide class of quasi-experimental treatment effect estimates using direct analogs of the features that make a randomized controlled trial transparent. The outcome weight measures how much each observation's outcome contributes to an estimate, and its sign defines effective treatment and control groups. We show estimates take a Wald form, equal to the weighted outcome difference between these groups divided by an effective first stage that quantifies the identifying treatment contrast. Our framework also allows researchers to assess covariate balance, examine how differences throughout the outcome distribution contribute to mean impacts, and identify influential observations. The treatment effect weight--the outcome weight times treatment--instead measures how much each observation's unobserved treatment effect contributes to an estimate, and exactly decomposes an estimate into subgroup-specific components. When the implied weighting of an estimate is undesirable, we show how to construct alternative estimates that satisfy researcher-specified criteria. We illustrate the framework with analyses of democracy and growth, Chinese import competition, and returns to schooling.