 We were looking at the lessons and gaps of the projects of the S&P project and we found some lessons and gaps which are actually linked. So we have, for example, the main thing that we need, that a project like this needs somebody, 100% who takes it, to take care and move the project. So that means that we need also human resources, but it needs some trade-off between coordination and harmonization and also mandating a bit the process. So there should be flexibility for regions, but not too much because that will just disperse us a bit. We also need, among ourselves, we need a common understanding of what we are doing, so the tools we are using, why we are using them. That's something we learned actually. What we learned also, we need to build things that is a knowledge sharing for the different partners, for farmers and local partners. Also considering the local calendar of farmers, so weeding, harvesting times. We need also, when we are thinking about developing a tool, we need to consider farmers' knowledge about everything so that we ask them questions that they can answer and that they can lost. Well, related to all of these, there are still some gaps. We need to develop a framework to analyze the data so that we can analyze it again keeping some flexibility between regions, but we need some framework at least. Also what we would need is, the thing is also, again, between regions we also need to build better links between, for example, models and regions, but also the communication and maybe resources, exchange of information and actors as well between regions, across regions. Instead of just regions, central office, CIA, the rest, but more communication between, among regions. What we also need to develop and related to the flexibility and not too much flexibility is a better work plan on a timetable hitting on ourselves. But also we need, the gaps I think we need to build a better idea of what capacity building we want to build. Well, capacity we want to build with the local farmers. And the most important thing is, okay, the research project goes until 2011. This big one. There is money for the third one, but we will see. But it's, okay, how are we going to disseminate information? What are farmers going to get out of this? Are we going to work on interventions in terms of community or policy making? And how farmers are going to get something out of this? And also related to that is that so far we have focused too much on the tools, so we are very efficient in going to the field, but we don't spend so much time in the field. So we might get wrong data because farmers just see us just getting three, four hours, getting a villa service, see you, goodbye. So we are missing a bit their knowledge and with that knowledge we can also build a better framework for analysis data. We can have a better access to information. And so that's a quite important one that we are missing. One thing I would add is I don't see any discussion of the process of developing this project and some of the iterations in terms of the thinking they went on through the process. Whether we could have arrived at where we are now more efficiently or whether we really needed to go through some of those taking four proposals and trying to harmonize them into a single proposal. Was that a useful process or was it a waste of time? Could we arrived at where we got to more effectively using a different process? Perhaps I could add that we have been talking very much about the process of the issues as they come. But whereas it is often more efficient if you start from the end. We do have our research questions which we sometimes remember to think of, but the steps before the research which would lead to the results, the analysis steps, we only have very rough ideas even now while we are developing the service. It is a common problem with all projects, but I think it is still a gap that it is more helpful if you have a clearer idea of the analysis steps at an earlier stage so that we can be more efficient in our data collection. There is always the question we collect so much data, it is too long, do we really need all of it? We cannot say without knowing more or more precisely what we will actually work on.