67

While an array of software offers to monitor the brands’ reputation on the web, we question the processes by which these dashboards actually aggregate and describe the judgments made by people on the web about companies and products. These tools address in fact a major methodological challenge: converting the “noise” into a “signal”, that is to say, to make intelligible preexisting, massive and heterogeneous verbal data. Through the study of French and American providers, we show how the way they are designed and sold make these dashboards permeable to the client’s epistemologies and cognitive frameworks. By requesting from the user constant arrangement and interpretation of algorithmic results, the technical and commercial procedures that come with the dashboards allow the matching between the client’s questions and the flow of online speech.