In a petrochemical factory, we have managed to collect short videos of subcontractors' industrial cleaning work. To carry out an ethnography of work in this factory, we were submitted to several series of control procedures. The videos we propose to analyse were collected soon after our arrival on the field. They were supposed to complete the design of the research, mostly based on direct observations. The regulatory framework “Seveso II seuil haut”, which is applied to this classified site for its risks, marked the negociation of the setting of the collection of these videos. As a result, a technician of the factory filmed for us the work of the present subcontractors. The technician made short films with a small compact camera. Through interactions with him and other agents about the setting of this collection, “work” was defined as our research subject. When capturing images, the technician made choices according to what he thought would interest us and to what he thought was important in this situation. Viewing the sequences, long after their making, has been source of information for ethnography: by highlighting technical objects, the content of the images has emphasized the first impressions of direct observations. It cast new light on the question of the definition of work: is work “located” in human beings or technical devices? The viewing of the collected videos enables the ethnographer to specify the definition of the object in contact with the field.