The aim of our paper is to analyse the determinants of the innovation propensity of the firm. Among the numerous works devoted to this subject, the interest of our research is, firstly, to use a direct measurement of innovation, instead of the usual proxies, as R&D expenditures and patents statistics, secondly, to emphasise the role of labour factor quality as a major determinant of innovation. We first build a definition of labour factor quality, based on a double dimension: individual skill level and functional distribution of jobs inside the firm. At the end we consider that each job category can be involved in the innovation process, at the different steps of it: conception, decision, implementation. To explain the innovation propensity at the firm level, our logit model takes into account four explanatory dimensions: the quality of labour factor employed inside the firm, the firm structural characteristics (as size, for instance), the sectoral market structures and, finally, the quality of labour factor employed inside the firm sector, as a proxy for the R&D spillover effect. We use some individual firms data, including a direct measurement of innovation, that distinguished between several types: radical vs. incremental and product vs. process vs. organisational innovation. The French food industries with its 500.000 employees and 42 sectors, mostly composed of small firms, are our empirical field. The results emphasise the influence of the usual firm structure variables. Firm size, particularly, is very clearly positively related to the innovation propensity. At the same time, some more original facts appears, such as the influence of firm status: after controlling the sectoral influence, co-operative firms seem to innovate less than private ones. Labour factor quality appears to play a very significant role by itself, but mostly, helps us to analyse and specify the influence of other variables on innovation. At the end, it shows that innovation is a multiphase process, and that the relative importance of each phase greatly depends on the kind of innovation that is considered. Conception is the most important phase in the radical innovation case, which greatly involves formally high-skilled job categories as R&D employees or engineers. At the same time, the implementation phase, which seems to be particularly important in the incremental innovation case, emphasises the role of the intermediate categories know-how. At the end we can say that small industrial firms appear to be less innovative for two reasons: the usual scale effect argument is correct only in the process innovation case in relation to the capital intensity level. In some other cases as radical innovation, small firms are less innovative because of their job structure and particularly because of the lack of formal scientific capabilities (as the R&D personnel''s one).