Big data and agriculture


Posted on September 1, 2016  /  0 Comments

Big_data_agricultureI was asked to make one point about the way forward at the closing session of the excellent e agriculture solutions forum organized by the FAO and ITU offices in Bangkok. Here is what I said (more or less, but this is the jist):

Big data in agriculture

We have come a long way from being fixated on radio as the be all and end all of ICTs in agriculture. We are fortunate to be living in an age when we can even take smartphones for granted in Myanmar, a country still listed as an LDC and one which went from 10 mobile connections per 100 people to over 80 in less than two years. Our own surveys (early 2015) showed that 63 percent of all mobile owners in Myanmar had smartphones, with more computing power than the computers we used just a decade ago. The mean price of a handset was USD87, with the largest number being in the USD 50 range.

Average monthly top-up per poor household was around USD 7 back in February 2015. Prices have come down since, so monthly spend should have decreased somewhat. We will have the results of the 2016 survey in a few months. More than 50 percent of mobile subscribers in Myanmar are data users. This leapfrogging indicates the conditions are right for e agriculture initiatives.

If we are looking forward, we have to now shift focus to the “I” in ICTs. We have to provide customized information at the right time to the participants in agricultural supply chains. If these conditions are satisfied, the debate about whether we need to give information and advice away for free or not will become irrelevant.

My expertise is not in agriculture. But my organization has been working on e agriculture since 2006. One of the many things I learned is the tremendous heterogeneity of information needs. What is relevant at planting time is not relevant at harvest time. And the area that we started in, fruits and vegetables, the variety is vast. What is relevant to a farmer growing chilies is not relevant to one growing cucumber. We’ve done systematic reviews of research on impacts of agricultural information services delivered over mobiles. One reason the results have been unimpressive is because they do not take this heterogeneity into account. But how, is the question.

In a number of instances, for example in discussions of traceability, we heard references to data analytics. But we did not get into too much detail.

Big data is what is being used by marketers to customize messages. Data from multiple streams are being aggregated and analyzed in order to better understand market segments and indeed specific prospects. Based on these understandings, customized messages are developed. Not fundamentally different from what is required in agriculture.

We at LIRNEasia are working on big data for public purposes. We work with pseudonymized call detail records from mobile networks. In Sri Lanka, this amounts to 100 million records in one day. Our work has not required market segmentation, but we know what can be done.

What we need to think about as we implement the national e-agriculture strategies and the international work supportive of them is how we can use big data to better understand the information needs of participants in agricultural supply chains. You know the costs of using surveys to learn about information needs. We have done surveys and will continue to do more. But what is attractive about big data is that the data are a by-product of something that is occurring in a computer-analyzable form. There is no cost to giving the information. There are no errors caused by faulty memory or bad survey design.

Data analytics are already being done for the high-end agricultural producers, as we heard. The kinds of platforms and services aimed at smallholders described in some presentations here indicate datafication is moving toward the smallholders as well. Greater focus on the problem of customization and the potential of big data will help us realize the objectives of e agriculture.

Image source.

Comments are closed.