The presentation will deal with three discussions : 1) Roles of agro-food MNCs in Africa ; 2) Normative and Conceptual challenges of sound corporate impact measurement for SDGs and 3) Methodological challenges of sound corporate impact measurement in the context of SDGs. When it comes to MNCs, two main development models deserve to be coupled (Paul and Barbato, 1985). The second one (North-South model) has 7 criticisms for the older (Economic Development model classique). We reframe these 7 criticisms about the agro-food MNCS in Africa. These MNCs therefore emerge as a strong contender for creating growth without causing harm. On the other hand, in the context of State shortcomings, the MNCs are increasingly called upon to play a positive role (Kork and van Tulder, 2010). From Kostova and Roth (2008, page 1001) MNCs « are embedded in multiple, fragmented, ill defined » institutional systems, and are proactive in setting institutions. This is especially the case for the agro-food MNCs operating in Africa. Thanks to primary data stemming from the tropical fresh fruits and vegetables industry in Africa, we report on the interventions of MNCs following two aspects. The first one is a continuum between setting private or for public use facilities. The second aspect contrasts the facilities created in spatial proximity or across several regions (distal). The proximal/distal aspect and the knock-on benefit upon the industry contribute to explain the rules of the collaboration of MNCs with various actors. About the interventions of MNCs, the sound corporate impact measurement regarding SDGs is essential, and raises normative and conceptual challenges. SDGs have been chosen by actors, and even ranked in a hierarchical order. Nevertheless, the scientific law (or the laws) is missing, which could bind together the different goals, make them compatible, clarify the role of MNCs, and eventually highlight the way to achieve them. In the absence of this theoretical sound frame, what to do? We provide an example of potential scientific framework. Regarding the conceptual challenge, the main issue seems to be : do we deal with performances (features of the MNC at one given moment) or do we deal with impacts (consequences) of one change in the MNC’s activity (Vanclay, 2003)? The link between the performances and the consequences are unclear (Jørgensen et al., 2010 ; Ardvisson et al., 2014). Moreover, the MNC can distort its activity in order to get good scores for the performances required by reporting (Paradeise, 2013). The MNC can also shift the focus away from the firm serious failure (Laufer, 2003) by reporting about some selected anecdotic but emotionally charged performances (Loeillet, 2013). Depending on this first main choice, other complementary concepts are necessary. What is the perimeter of the assessed system? Is it the local agency of the MNC itself, the supply chain, the organizations belonging to the life cycle of the product (Dreyer et al. 2006), or the strategic network (Brandenburger and Nalebuff, 1996) including the firms in coopetition with the central player, and suffering or taking advantage from the change? What are the spatial and temporal perimeters of the effects under scrutiny? It is the whole population or one part of the recipient place where the central firm (Gereffi et al. 2005) is set up? It is the population of all the developing countries involved in the life cycle, or from all the countries (developed included)? We will suggest some insights out from our experience of social life cycle assessment (Macombe, 2013). Among the methodological (in the sense of collect, analysis and interpretation of data) challenges, scarcity of data is noticeable (but this challenge can be met in agro-food sector from the reconstitution of the agricultural technical processes). Nevertheless, two other challenges are prevalent : i) how to succeed in getting the assessment to be done by “real” external third party ? Usually, the auditors are forming an « epistemic community ». They are at the same judges and judged in the matter (Fouilleux, 2013; Falque, 2014) ; and ii) designing the fair baselines and perimeters, in order the results to be attributable to the actions of the MNC. In this regard, we highlight the advantages to choose the approach by impacts.