Survey research is used to investigate a variety of different constructs, such as beliefs, behavioural intentions, perceptions, preferences and so on. Despite the wide range of constructs studied by social scientists, the ordinal answer format tends to be used across the majority of survey research studies. We challenge this standard approach in survey research by hypothesizing that the ordinal answer format is not optimal under all circumstances. Instead, we propose that the suitability of answer formats depends on the construct measured. We conduct a repeat measurement study using binary, ordinal and metric answer formats measuring two different constructs: beliefs and behavioural intentions. A clear interaction effect between answer formats and constructs is revealed. This supports the notion that no single answer format is optimal for all research problems, but that some constructs are naturally more suitable for certain answer formats than others. These findings call for increased use of pre-studies to determine the optimal answer format before fieldwork is conducted rather than relying on standard answer formats.