 Agricultural research has several goals. One is to boost food production. Another is to improve the income and livelihoods of farmers. Many research projects draw a direct line between the two, then try to find a way to boost productivity in the hopes that farmers will produce more and become better off. Crop breeders try to develop varieties that yield more. Pest researchers try to find a way to control an insect that damages the crop. Livestock scientists look for vaccines to control diseases or a better way of keeping animals. It is the researchers who decide what the problem is. They come up with a new technology and roll it out to farmers. They expect that the farmers will seize the opportunity and adopt the new technology straight away. But it often does not work out that way. The researchers are disappointed to find that few farmers actually adopt the new ideas. Low adoption means few gains in productivity, little extra output and no big change in farmers' income. MilkIT, a project managed by the International Livestock Research Institute, tested a different approach. This project worked to improve dairy in India and Tanzania. Instead of having the researchers decide the priorities, it asked the milk producers and other people interested in dairying what they thought were the main issues. To do this, the project set up innovation platforms. Groups of people who were interested in the same thing, in this case milk. They were dairy farmers, traders, extension workers, input suppliers, people at various stages in the value chain. Each person brings their own skills and ideas to identify the common problems and find ways of solving them. The innovation platforms in India focus first on linking the farmers to the market so they could sell their milk. The farmers had complained that they couldn't sell the milk they produced. So the innovation platforms helped them organize groups of farmers to supply milk to nearby towns through self-help groups and cooperatives. They set up collection centres and employed people to test and weigh the milk and bulk it and arrange for the vehicles to transport the milk to town. Cooperative members even set up a shop in town to sell the milk. The prospect of earning money gave the producers a strong incentive to increase their sales. They started adopting all kinds of new production methods. They bought feeding troughs and choppers so they could make the most of their feed. They planted new grasses to produce more fodder. They brought improved animals that would produce more milk. They vaccinated their cows against disease. With markets in place, the innovation platform helped its members identify and overcome different problems. In Tanzania, the innovation platform members wanted to focus more on fodder production. The platforms looked for new varieties of fodder grasses. They helped farmers fence off land to use for grazing in the dry season. More feed meant more milk. But the effect was not as big as in India. Perhaps because the market linkages were not yet in place. So what can we conclude? What should a research or development organisation do? In India, focusing on the market was fast and efficient. It had a big effect on milk production but it was unpredictable. It was impossible to predict beforehand what solutions might be needed. The members of innovation platforms decided on the interventions. They, not researchers, drove the process. The innovation platforms in Tanzania used a more traditional approach focusing on productivity. The gains were slower and fewer farmers took up the new technologies. Only after some time did private processes take an interest in the increased local milk production. And they were concerned that quantities were small and could not be sustained year-round. So there is no right approach. But on balance, sorting out milk marketing early on can really stimulate farmers' interest in boosting production. If they can earn more by selling more milk, farmers have an interest in feeding their cows better. An assured market means they will want to invest in productivity. They will then start demanding new techniques and research and development organisations have to be ready to listen to them and to find the solutions. Other innovations will tend to follow on without much need for a push. Innovation platforms which bring together farmers, extension workers, traders, processors, policy makers and researchers are a good way to discuss these issues and find out what the priorities are. Innovation platforms should not be used just to push solutions that the researchers or policy makers think will work. Instead, the farmers and traders should do the talking and the researchers and policy makers should spend more time listening. Only then will they be able to find solutions that the farmers and traders will actually use.