Increasing both survey completion rates and data quality remains an important topic for fields as diverse as sociology, marketing, medicine and history. Thousands of studies have made response quality their central topic of examination, but their focus has largely been to measure response bias through the comparison of early–late wave responses. In this study, an innovative online field experiment tests a two-staged highly interesting question to produce an 8% better survey completion rate and to change sample representativeness by 12% over a usual one-stage highly interesting question appearing at the beginning of the questionnaire. In addition to these substantive findings, a distributional and probability analysis is developed that further refines methods for identifying the extent of non-response bias.