This article is concerned with the meanings that employees in industry attribute to representations of data and the contingencies of these meanings in context. Our primary concern is to more precisely characterize how the context of the industrial process is constitutive of the meaning of graphs of data derived from this process. We draw on data from a variety of sources, including ethnographic studies of workplaces and reflections on the design of prototype learning activities, supplemented by insights obtained from trying out these activities with a range of employees. The core of this article addresses how different groups of employees react to graphs used as part of statistical process control, focusing on the meanings they ascribe to mean, variation, target, specification, trend, and scale as depicted in the graphs. Using the notion of boundary crossing, we try to characterize a method that helps employees to communicate about graphs and come to data-informed decisions. (Contains 5 footnotes, 5 figures, and 2 tables.)