Research methods

In a randomized policy-message experiment, some participants assigned to receive the message never see it. What remains true about the original random assignment, and what is a key caution?