In the mid 1980s, following the first stages of the decentralisation process initiated in 1982, certain French administrative regions started acquiring tools for the observation of the training-employment relationship in order to prepare for their new responsibilities in the areas of education, training and economic development. In so doing, the government reasserted its ‘determination' to anticipate in order to take action (not without echoes of the Keynesian state inseparable from national statistics), but this was done in a way which remained compatible with the emergence of a decentralised state: by establishing a framework rather than laying down rules. A circular dated 16 February 1988 thus gave the OREFs a common goal – developing diagnostic and forecasting tools to be shared by the regional actors (Bertrand et al. 2003) – but the definition of the specific institutional form and methods was left up to the observatories themselves. Each French region thus came to choose the kind of information produced and the ways of using it in the context of regional employment and training policies (Pascaud 1993). In the two regions under consideration here, the initiatives were quite distinct: in Rhône-Alpes, they took the form of an association, and in PACA, that of a structure within the Regional Council, although this was funded through the State-Region Planning contract. The implication of the OREFs in the specific regional histories of the training-employment relationship was all the more profound because the 1988 circular pragmatically tied their creation to the ‘multi-annual contractualisation' of public policy, as well as the ‘continuity of the transfer of authority over initial and continuing education and training to the Regional Councils. These two observatories have thus been studied here from a socio-historical perspective in order to bring out the configurations of regional actors taking part in their creation and subsequent development. To what extent do interactions between those involved in the fields of employment and training (i.e., between organised ‘interest groups' including the devolved state services, local authorities, social partners etc.) have an impact on the modes of production and use of the information produced and mobilised by the two OREFs under study? In the following presentation, we shall examine the changing content of the tools these OREFs develop in relation to the ‘mandates' they take on and the autonomy they acquire in view of (possibly) gaining the status of expert institutions. Our hypothesis is that the statistical apparatus of the regional training-employment relationship depends on an institutionalisation over the medium term, marked by technical innovations, the ‘sedimentation' of numerous tools and the restructuring of their uses. Ultimately, we would maintain that the histories of these two observatories entail successive compromises between the heritage of the French-style ‘planning state' of the 1960s and 1970s and the emergence of territorial expertise networks in the context of a decentralised state. To shed light on these regional encounters between two statistics systems – training and employment – we shall examine the production of these observatories in terms of three subject fields: sectors and occupations, territories and individual pathways. More broadly, our approach is based on an analytical framework which considers the regulation of public policy as the resultant of interactions between three determinant elements, namely institutions, ideas and interest groups – ‘the 3i'. The OREFs redefine their missions and activities in response to this interplay, while periodically coming up against difficulties of a technical, organisational or political nature. The periodisation varies somewhat between the two OREFs, but not enough to warrant separate treatments; we shall thus deal with each period in terms of both OREFs.