‘Big data’ has great unrealised potential in most parts of the agricultural value-chain. We can divide that data up into several categories, each with its own good and bad points. Starting in 1972, the US Landsat program collected the original ‘big data’, but capability to perform meaningful analysis of the photos remained very expensive until recently. Other flows have begun in the past few decades, from private satellites, point-of-sale systems, land-based sensors, and aerial drones. Unlike Landsat, the various newer sources have different ownership statuses. Globally, most smallholders don’t generate the revenue to pay for any of the various proprietary data sources or analysis. But we see significant value in the application of machine learning/’big data’ techniques to publicly available satellite and other sources. Advances in information technology allow us to disseminate good-quality yield, drought, and other analyses at a much lower cost than previously. As a result, relatively small external contributions can bring the established benefits of modern modelling expertise to a hugely broader and more diverse audience.