While collaborative (or joint) working between social services and primary healthcare continues to rise up the policy agenda, current policy is not based on sound evidence of benefit to either patients or the wider community. Both sets of practitioners report benefits for their own work from adopting new arrangements for collaboration. The underlying assumption behind much of this activity is that a greater degree of integration provides benefits to both users and their carers, a perspective that at times obscures the issue of resource availability, especially in the form of practical community services such as district nursing and home help. At the present time there is insufficient evidence to demonstrate that formal arrangements for collaborative working (CW) are better than those forged informally between committed individuals or teams. Furthermore, arrangements for CW have not hitherto been widely evaluated in systematic studies with a comparative design and focus on outcomes for users and carers rather than on processes. In this paper we propose a number of process measures for future evaluation of CW: (1) study populations must be comparable; (2) details of how services are actually delivered must be obtained and colocation should not be assumed to mean collaboration; (3) care packages in areas of comparable resources should be examined; (4) both destinational outcomes and user-defined evaluations of benefit should be considered; (5) possible disadvantages of integrated care also need to be actively considered; (6) evaluations should include an economic analysis. Those implementing new policies in Primary Care Trusts have, at present, little sound evidence to guide them in their innovative work. However, they should take the opportunity to rigorously test the advantages and disadvantages of collaboration.