Since the time Glific has been launched, many NGOs have started their journey with pilots and used their learning to make their actual program launch a success. Digital Green is one of those NGOs that made sure to use Glific at its best to make an impact on its beneficiaries.

With a mission empowering smallholder farmers to lift themselves out of poverty by leveraging the collective power of technology and grassroots-level partnerships, Digital Green started their journey with Glific with a specific goal, they are looking to engage smallholder Chilli farmers in Andhra Pradesh by sending timely advisory information that will help them with preventive steps from Leaf Curl, a plantation disease, and curation steps if their crop is already manifested with the disease during the crop cycle.

It was a year-long journey starting in 2020 to mid of 2021, The DG team was very cooperative and always ready to utilize the learning from the pilot and make required adjustments to the actual program launch.

(Source: Digital Green)

In the early stage of the pilot, they figured out that farmers were finding it difficult to type in local language to their queries or even for a simple question like the name of their village. 

The farmers were comfortable with saying the answers instead of typing them and that’s where the idea of voice bot got germinated. 

Digital Green collaborated with NavanaTech and integrated it into Glific to create a platform that allows farmers to engage with Chatbot through voice notes.

(Source: Digital Green)

In Jan 2022, DG ended the program as the crop season ended too and started getting ready for the next phase around March 2022 when the next crop season will start. 

The data collected through the program has a lot of insights on the behavior of the farmers and how they were interacting with the Glific Chat/Voice bot.

Interaction with farmers

DG has enrolled 199 farmers on this program out of which 82% have joined through chatbot and others through IVR. 

(Source: Digital Green)

For better and effective engagement with 199 farmers, DG used visual content as well. They figured out relatable visual content that helped farmers to understand the activities better.

(Source: Digital Green)

In order to increase the engagement, Digital Green designed targeted traffic generation ads to promote the program in the farmer’s community

(Source: Digital Green)

Farmers also share images of the crop as part of the program. The farmers were able to understand the activities and were participating to get benefits from the program.

(Source: Digital Green)

Learnings

With all the data collected from the pilot and actual program, DG accumulated these learnings that helped them to understand their target audience a bit better and something to utilize in the upcoming programs as well.

1. Voice was seen as a preferred way of responding to messages when farmers were asked to respond. This also indicates that farmers read/hear and act on the instructions provided in the messages. Typing was the least popular mode of responding to messages. [Please note that this may not be applicable in general to all types of users other than farmers]

2. Farmers who were successful in responding by voice the first time were not hesitant in responding using voice in subsequent messages too. Farmers who were unsuccessful in giving a voice response even after multiple attempts on the other hand preferred using buttons in subsequent messages.

3. Choice of keywords to initiate that first conversation from the farmer’s side should be done carefully. Our bot used a simple word “Hi” for this purpose but it was also a word that farmers generally use in their conversations, which led to the flow getting re-started for some farmers.

4. One of the reasons reported by farmers for not responding to messages was that they were busy at that time leading to them forgetting to respond. For such messages specifically where input from the farmer is mandatory to move ahead in the flow, an additional nudge should be sent after a span of 3-4 days for a recall

5. Schedule the push messages from the bot to the farmers around the time when they are generally active on their phones. While the field team can suggest additionally the responses received by the bot from the farmers can help refine these time slots

(Source: Digital Green)

6. IVR was not a good channel for onboarding new farmers in AP. The solution should allow for a hybrid way of accepting responses – buttons, text, voice. There were 11 users who were quite active in the flow but did not use voice even once to respond

DG Program details in-depth

All the findings were only possible with data analytics and effective program design. They identify the data points in the program design phase which helped them to get an in-depth analysis of the program and the stages it includes. 

Overview of the program (Source: Digital Green)

The program has 4 main query blocks for the user engagement model

1. Symptoms / Weather query
2. Crop Stage Query
3. Village name query
4. Adoption query

With the identification of these data points, DG generated various analytics and reporting on the data collected through the farmers. The data helped them to understand the behavior of the farmers and to make their upcoming program more effective and impactful

Voice responses received at each node

(Source: Digital Green)

User engagement at each node

(Source: Digital Green)

Symptoms/weather query node 

(Source: Digital Green)

Cropstage query node 

(Source: Digital Green)

Village name query node 

(Source: Digital Green)

Adoption query node

(Source: Digital Green)

Learnings specific to the program

1. If the primary focus of the solution is a theme that is incidental in nature like pest issue in a crop then it is a good strategy to include more services that can cater to a wider group of farmers. Our decision of including preventive advisory and weather information into the service offering was a good one. It helped us engage with additional 26% of farmers

2. Farmers who see a direct benefit for themselves from the service are more likely to engage with the service. We saw farmers in the curative flow engaging and responding more than farmers in the preventive flow. Voice was seen as a preferred way of responding to messages when farmers were asked to respond. This also indicates that farmers read/hear and act on the instructions provided in the messages.

3. As the farmer on-boards the service, the system should not bombard her with messages. We were sending 5 messages one after the other(Welcome, Leaf Curl symptoms, demo video, audio for receiving a response, gif on using voice note feature) the moment the user joined, this might have resulted in key messages getting overlooked leading to receiving lower response using Voice from farmers.

4. When designing Ad campaigns for onboarding farmers the messaging should be specific in its offering and not very subjective. We received a better response when we ran the FB ad with a direct message on Leaf Curl or Chilli advisory this Kharif as compared to the campaign which said it would increase productivity and reduce the cost of cultivation.

What’s Next?

It’s been a great experience working with Digital Green on this program and all these learning, information, and experiences will be utilised in the next upcoming program. 

Glific team has also shared their experience working with Digital Green in these blogs. Please do give them a read as well.

• In conversation with Digital Green | GlificTalks#1 – Abhishek Sharma
• Digital Green’s voice-based chatbot for Chilli crop farmers – Mohit Sharma
• Digital Green: Chatbot based farming advisor – Akhilesh Negi
• Reaching farmers directly, scaling reach, reducing time to get analytics from months to seconds – Abhishek Sharma