Smartphones have a large potential for improving data collection by using research apps and collecting sensor data. This brings opportunities to enhance or extend measurement and to simplify the response task for respondents. Sensor data can (partly) replace survey questions, and these sensors potentially generate better data than respondents can provide themselves. This seems very promising, but many methodological questions arise related to representation and measurement in smartphone surveys; are respondent willing and able to participate and share sensor data, and how useful are the additional data collected via sensors and apps? In this dissertation we investigated the effect of smartphone surveys in terms of reducing (or enlarging) error components. These errors can be divided into errors in the measurement and the representation. In Chapter 2 we perform a systematic review and meta-analysis to investigate how to improve the effectiveness of the consent to data linkage question. In Chapter 3 we study nonresponse and nonresponse bias in the smartphone-only version of the Dutch Time Use Survey (TUS). In Chapter 4 we focus on measurement error when collecting GPS data in a smartphone survey. In Chapter 5 we investigate both representation and measurement in an innovative and experimental study on the use of sensor data in fitness and health research.