HeadCrowd is a collaboration between the School of Mathematical and Computer Sciences and the School of Textiles and Design at Heriot-Watt University. It investigates how rich web and mobile applications may be employed to provide designers with near instantaneous and highly visual feedback from thousands of potential customers, or crowds. We are exploring the use of state of the art rich media applications to add quantity, speed and statistical accuracy to the study of semiotics, and the use of visuals in fashion as communication. The project seeks to add to participatory design and market intelligence processes by enabling rapid and iterated co-design cycles between crowds and designers based on visual forms of communication so as to mirror the highly visual nature of fashion design inspiration. Such a scheme shows applications for sustainability in fashion if it can give crowds a concrete sense of ownership of the design process and provide enthusiastic target markets, thereby offering potential to significantly reduce the risks of producing unwanted product. The paper provides an analysis of prior knowledge before describing the first two stages of the project, in which a pilot browser has been constructed that allows observers to navigate a vocabulary of 500 images which have been ordered into 48 similarity stacks using a mixture of human and crowd sourced sorting techniques. A first test involved the presentation of 20 terms to observers and asking them to choose 3 images from the browser to represent each term. Analysis of the resulting pilot data has given insights into the communicative certainty that a selection of 3 images from a vocabulary of 500 can provide for certain types of terms, and amongst certain groups of testers. It has also prompted deeper analysis of the pilot browser. To put the communicative value of visual feedback to the test, the current research phase is preparing the reverse experiment of asking a fresh cohort of participants to associate images back to the original terms, and various interfaces are currently being constructed to facilitate the presentation of visual choices from phase 1. The similarity relationship between test images is investigated and visualized before a case is made for comparative experiments of raw selection data and versions of visual summaries in this second research phase, in order to test which way of data presentations best convey the intended visual communication.