Match your cases to your target group

Keep the group you want to learn about separate from the cases you actually collect data from. You’ll reason through how a question about “all commuters” might become a table of 200 surveyed commuters.