When relational data are exported, the results are flattened in a way that makes data analysis unbearably complicated. Suppose a question asks the user to enter up to 5 foods they ate, and the following page pipes those responses into a question which asks the respondent to indicate if anyone in their group at the same food (yes, no, DK). Currently, the export will create one column for **for every food item ever entered in the survey**. This means that if I have 100 respondents and each of them enters a different food, I am going to end up with up to 500 columns for each food ever entered (assuming the values are unique). If I have another question that asks what day the food was consumed, there will be another 100+ columns in the export representing the response for each food item. Is it not possible to set up the export of relational data in such a way that there are only 5 column headers, i.e, FOOD1, FOOD2, FOOD3, FOOD4, FOOD5, with a value for each food type entered, and then another 5 columns labeled "Group Meal1" through "GroupMeal5" (or however the user specifies the field should be labeled) with a yes/no/DK response recorded in each cell? There should only be a single row per respondent. Ideally users should be able to specify the variable names for the rows/columns ahead of time, during the survey definition/editing stage, before defining the criteria for data export -- otherwise the user would have to redefine the row/column labels each time they are ready to export the data (not desirable!). The current export functionality is also messy with matrix-type questions that allow the user to combine free-text, radio, and checkbox types into a single question.