Job exposure matrices (JEM) are designed to link information on occupation with information on exposure to specific workplace hazards. In spite of some limitations, JEM are particularly useful in large retrospective epidemiological studies. The development of JEM designed for a company or an industrial sector have a more specific field of application than JEM used in population-based studies, and can therefore be based on a more detailed classification of occupations and better exposure information. This article reviews on-going research related to the design of JEM in support of industry-based studies. The review covers design aspects related to the structure of JEM, and specially the four main axes that may produce differences on exposure characteristics: agent (exposure), job, time and place. Attention has also been given to the different sources of information on exposure, the characteristics of the exposure and the validation of exposure estimates.