The observation that politicians are increasingly dependent on opinion polls is a common thread in public debates as well as in the scientific literature (Druckman &amp; Jacobs 2015; Belot 2012; Kruke 2012). While research on the way policymakers make use of surveyed public opinion is bourgeoning, their conclusions vary widely. On the one hand, some analyzes suggest that governments use polls above all for persuasion purposes (Page 2006; Jacobs and Shapiro, 2002; Heith, 2004; Eisinger, 2003; Towle, 2004). On the other hand, responsiveness studies affirm that surveyed public opinion has at least some influence on public policy (Stimson 2007; Manza and Cook 2002; Page 2002; Glynn et al. 2004, Burstein 2010).We believe that part of the difficulty to assess the role of polls in the decision-making process lies first in the fact that most studies do not have access to governmental polls and tend to consider them from the tip of the iceberg, questions and results quoted in speeches or media. As recently shown, these results are biased (Gahner Larsen &amp; Fazekas, 2021) and mostly hide what is at work when those who govern commission polls. It further lies in the fact that most studies consider questionnaires as a whole, as if they were commissioned with one and only purpose.Having systematic access to all polls commissioned by the German government and the European Commission, our research aims at disentangling which goals those who govern pursue when commissioning polls by systematically coding each question. We argue that, depending on the type of questions, the logic might not be the same and that the systematic analysis of the types of questions asked can deliver first insights in their intended use by governments. The only existing classification of question types was proposed by Druckman and Jacobs (2015) who distinguish seven types of questions without however applying it empirically. We propose a new typology of question types distinguishing 5 categories (preferences, evaluation public policy, evaluation situation, evaluation political life, and socio-demographics) and 25 subcategories. The coding of 15 supplementary variables allows considering further information on the character of the questions. Applying this typology to all questions (about 20 000) commissioned by the German government and the European commission between 2013 and 2017 allows us to fully describe their polling practices, finding patterns and regularities as well as critical cases depending on the commissioning actor, electoral and policy cycle contexts, issues and issues salience. It allows us to go deeper into the understanding of the role of polls in the decision making process and to develop more intricate hypotheses for further researches.< Réduire