In many studies, consumer preferences are determined by using direct surveys. For this method social desirability is problematic. This leads to the effect that participants answer in a way that they perceive as desired by society. This leads to the stated importance of certain features in these studies not being reflected in real purchasing decisions. Therefore, the aim of the study is to compare consumer preferences measured by a quasi-experiment to those quantified by direct questions. Another objective is to quantify the part-worth utilities of product characteristics such as origin, price and food labels. Part-worth utilities are estimated on an interval scale with an arbitrary origin and are a measure for preferences. The real purchasing situation was simulated in a quasi-experiment using a choice-based conjoint analysis. The part-worth utilities were then compared with the results of a conventional preference assessment (Likert scale). For this purpose, 645 consumers from all over Germany were surveyed in 2014. The participants were on average 44 years old and 63% were women. The results of the conjoint analysis report the highest part-worth utility (2.853) for the lowest price (1.49€), followed by the characteristic “grown locally” (2.157). For the labels, the German organic label shows the highest part-worth utility (0.785) followed by Fairtrade/“A heart for the producer” (0.200). It is noticeable that the carbon footprint labels have negative part-worth utilities compared to tomatoes without a label (−0.130 with CO2 indication, −0.186 without CO2 indication). The price is ranked 12th in the importance of the characteristics of purchasing tomatoes in the survey with a Likert scale, whereas it is first in the evaluation of the quasi-experiment (conjoint analysis), which supports the assumption of a social desirability bias.