We analyse the results of experiments on questionnaire design and interview mode in the first four waves (2008-11) of the UK Understanding Society Innovation Panel survey. The randomised experiments relate to job, health, income, leisure and overall life-satisfaction questions and vary the labeling of response scales, mode of interviewing, and location of questions within the interview. We find significant evidence of an influence of interview mode and question design on the distribution of reported satisfaction measures, particularly for women. Results from the sort of conditional modeling used to address real research questions appear less vulnerable to design influences.