Professional judgement is a key component of questionnaire development, subjective in nature and rarely reported in this context. It is required in dietary questionnaires to delimit the size while retaining quality of the data. The objective of this study was to describe the nature and extent of professional judgement involved in developing a food database to include in a web-based self-administered dietary assessment. Professional judgement was applied in tandem with a stepwise statistical analysis of hierarchically reported foods in the Australian National Nutrition Survey (NNS95). Statistical analyses determined foods commonly consumed and eaten together and three different forms of cluster analysis were then used to group foods that were most similar in macronutrient content. Professional judgement was required to interpret these groupings and determine the most suitable clustering technique. Face validity of the resulting food groups was determined by recognition of the food name by experienced dietitians, as usually reported in a diet history interview. Applying professional judgement to differentiate between foods after the cluster analysis resulted in an increase from 370 to 501 food groups. A final three-level hierarchy of 19, 103 and 422 groups in the new database compared with 21, 106 and 370 groups of NNS95 was developed. The use of professional judgement in database development is an important step when they are to be used in self-administered assessments. It ensures foods are not only nutritionally appropriate but also conceptually appropriate for recognition by a layperson.