This paper estimates causal effects of two Polish active labor market policies - Training and Intervention Works - on employment probabilities. Using data from the 18th wave of the Polish Labor Force Survey we discuss three stages of an appropriately designed matching procedure and demonstrate how the method succeeds in balancing relevant covariates. The validity of this approach is illustrated using the estimated propensity score as a summary measure of balance. We implement a conditional difference-in-differences estimator of treatment effects based on individual trinomial sequences of pre-treatment labor market status. Our findings suggest that Training raises employment probability, while Intervention Works seems to lead to a negative treatment effect for men. Furthermore, we find that appropriate subdivision of the matched sample for conditional treatment effect estimation can add considerable insight to the interpretation of results.