Reaching farmers directly, scaling reach, reducing time to get analytics from months to seconds

Abhishek Sharma

NOVEMBER 08, 2021


This case study has been prepared in conversation with Erica Arya, the program manager at Digital Green.

After their two chatbot programs, Digital Green has been working on another, more innovative one. They see great potential in automating the parts of programs that are repetitive, routine and don’t need manual intervention. They have also explored multiple channels to strengthen their program. Along with in-person group training, they’ve added a more direct channel to reach farmers on their WhatsApp via the chatbot. They believe the use of technology and solutions like chatbot are the right way to scale their reach. Their journey from the discovery to successfully deploying chatbot wasn’t any shortcut. It’s surely an insight for anyone interested.

About the program with farmers

Digital Green is running a voicebot with Glific, which targets Chilli farmers in Andhra Pradesh. We want to help them solve the leaf curl issue in the chilli crop. Leaf curl is a pest infestation that can happen multiple times to the same crop. We send advice to farmers so that they can take better care of their crop and reduce instances of pest Digital Green is running a voicebot with Glific, which targets Chilli farmers in Andhra Pradesh. They want to help the farmers solve the leaf curl issue in the chilli crop. Leaf curl is a pest infestation that can happen multiple times to the same crop. Advice is sent to farmers so that they can take better care of their crop and reduce instances of pest infestation. Since the infestation recurs, DG checks with the farmer every 15 days if they are seeing the issue. Besides sharing advice on the infestation, they also share weather advisory to the farmers. Farmers just need to let DG know their village name and then location specific weather forecasts are shared with them for the week. DG believes that better information is critical for farming. The duration of the bot is 90 days. “We take information from the farmers about their crop stage to send timely, personalised, advisory to the farmer,” said Erica. “Glific, as the tech partner is helping to build the use case as per the workflow designed by us.”

After trying a couple of chatbots, they’ve designed one using voice. The farmers can respond by recording and sending their message over whatsapp. Glific then uses transcription services by Navanatech to understand the response and send the farmers the next message. In the previous chatbot experiments Digital Green found that farmers struggled with typing. Experiments with them showed the use of photos and videos, but less conversation was happening on the chatbot especially for not so digitally savvy users who preferred local dialect. Typing was a barrier and hence they are testing voice to see if that is a preferred mode of communication. So far there has been a mixed response. In some cases, voice was preferred for questions that didn’t have predictive answers but needed their response such as asking their village name. While in other places, interactive buttons(not exactly typing), for users to simply tap and indicate their response was preferred.

Behind using WhatsApp as a channel

It is interesting how DG reached chatbots as a solution. It has taken a fair bit of experimentation, exploration, trials to make it a part of the program delivery. For more than a decade they have used in-person video screenings as the main channel to disseminate advice to their communities. Feedback on the services provided by Digital Green was getting missed in this process. There was a need to get more insights from the farmers to improve the services. During the video screen in a group, there was a fixed topic which all the attendees were required to watch. For a group consisting of upto 15 women, not all of them connected to the same topic. And when they didn’t find the screening relevant for them, they would disconnect and lose interest. Retention and application of learning in such cases is not that great. Although there were benefits of group setting such as peer learning, there was a need to make the advisory more effective. That’s when a direct channel that completed the group learning was added. This brought personalised information directed to specific needs of farmers, which was missing earlier in the group setting.

Sometime in 2019, the DG team went down to Bangalore and performed a scoping study to understand how their users were using smartphones. Without a specific motive, they wanted to understand the state of increasing smartphone usage in the country. They wanted to bring smartphones in their solutions strategy. After spending days with the farmers community, conducting interviews, learning how phones were used, the kind of content that was being consumed, they came up with some findings. WhatsApp, YouTube, and Facebook were common in all of their phones. WhatsApp was used extensively to share photographs, take video calls or forward messages. However, none of these channels were being used for the purpose of learning. No important information was being exchanged in groups. That’s when DG tried to introduce a more productive use of WhatsApp in the farmer community. It began as a manually managed group of 7-8 farmers which organically grew to 150 farmers. The group was active every day for 2-3 months of trial.

Motivation for the WhatsApp chatbot

It has been more than a year that DG is using the chatbot. What motivated them was automation of the routine, regular conversations, an easy to use platform that enabled drag and drop methods without relying on the technical team. They realised that they didn’t need a lot of technical expertise to run the program in this manner. They also discovered the wide variety of analytics and reports that were made available. This opened insights on how the users were interacting. Though technology was available, they began with finding a fitment for their use case. “We never realised that our use case was different from the general use cases that the bot is used for by organizations or by platform providers” said Erica about their process of getting started. She explained that it is better for organisations to come up with their requirements first, understand all limitations of the channel, and discuss it early with the solution providers. Discussing the requirements with the providers can help in getting ready to adopt the new technology. Erica also explained how a lot of details go into building use cases such as availability of all languages, or the automation part. The solution providers with their experience of working with other organisations, will be able to highlight anything critical, and provide solutions for the use case.

Preparing for the WhatsApp chatbot

There were not many products for the social sector at the time DG began their program, and they were eager to get started. Hence they went ahead with Haptik, to not waste more time in exploration. As a part of preparation, they also understood the various cost components – fixed and variable costs. They were more focused on exploring a new innovative channel than worrying much about cost-effectiveness at the start. There were many variables at the start, and many things to figure out, that cost-effectiveness couldn’t really be one of them. There has to be a sense of freedom, creativity, fearlessness about trying out new technologies and methods to run the programs. They are convinced that technology is the way to go when thinking about scale and hence there has to be exploration and innovation involved in the process. After getting a good handle of the technology, they will bring back attention to cost-effectiveness also. For organisations getting started, Erica suggests giving the channel a try even with a pilot, a manual form of communication. There’s a lot of discovery in the process of running a trial with a small sample of 50 people. One gets the lay of the land before thinking about optimizations in their processes, costs or skills.

Activities to ensure success

Digital Green also runs parallel activities to ensure the success of the chatbot. They periodically conduct feedback calls with the farmers to learn about their experiences. Some of the things they want to understand are the challenges farmers face and their rating on the program impact. They segregate the data between high frequency users – those who respond to most of the messages, and low frequency users – those who joined in but remained silent throughout. Feedback calls with the farmers help in improving the chatbot experience. 

Before the chatbot, DG’s intervention with farmers happened through extension workers. DG wasn’t directly interacting with the farmers but that changed through the chatbot. There’s a greater insight into how the farmers are engaging with the services. The platform makes analytics available at runtime so DG was able to know what worked and what didn’t. During the in-person method, it took 1-2 months to get the analytics and reports because the data would get collected and then digitized. Implementing any modifications and changes from the learnings would take a lot of time. On the other hand, they were quickly able to decide if chatbot was the right way to go because all the decision-making information, and feedback was readily available. On this Erica said “we are able to see user conversations, go through each of them, and understand the interest of our farmers. We don’t have to go out to ask for various information from anyone. This was missing previously.”

The transition and transformation of programs at Digital Green has gone through a lot. From discovering the user of WhatsApp among their community, to running pilots and manually moderated groups with a few hundred users, to deploying chatbots and now evolving that into a voice + text bot. It only opens up more possibilities and ways to deliver programs effectively. We’re excited to see how it will evolve and to partner with them during their process of further innovation.

Leave a Reply

%d bloggers like this: