Agro360 Bangla-Dialect A/B test
BAN -19 -1545Last modified on July 25th, 2025 at 6:04 am
-
Abstract
As part of the Agriculture 360 (Agro360) program, mPower sends text messages with crop management recommendations and weather alerts to farmers to improve the efficiency of their crops’ cultivation and promote practices that mitigate the impact of climate change on their crops.
The objective of this experiment was to understand farmers’ engagement with the service when the voice messages are delivered in their local dialect, compared to the traditional Bangla messages that have been sent to farmers during the previous seasons. The results show that farmers engage more with the voice message when it is sent in their local dialect, as reflected in the listening rate and probability of listening to the full voice message. -
Status
Completed
-
Start date
Q1 Jan 2019
-
Experiment Location
Bangladesh
-
Partner Organization
mPower
-
Agricultural season
Rabi
-
Experiment type
A/B test
-
Sample frame / target population
Mungbean and watermelon farmers registered in Agro360 and assigned to receive voice recommendations
-
Sample size
4,518
-
Outcome type
Service engagement
-
Mode of data collection
PxD administrative data
-
Research question(s)/hypotheses
Sending voice messages in local dialects would increase farmer's engagement with the service.
-
Research theme
Agricultural management advice, Message narration
-
Research design notes
The study randomized 4,518 farmers into two groups to evaluate the impact of language on engagement with voice message advisory services. One group (2,360 farmers) received messages in the local dialect, while the other group (2,158 farmers) received messages in standard Bangla. Randomization was stratified by farmer union, crop type (mungbean or watermelon), and planting date to ensure balance in key characteristics across treatment arms.
Because watermelon farmers joined the service later, they received fewer total messages on average.
The random assignment of treatment ensures that the different farmer groups have similar characteristics before the start of the experiment, therefore any changes observed in their pickup and listening rates can be attributed to being caused by treatment status.
-
Results
The results show that farmers engage more with the voice message when it is sent in their local dialect. Receiving a voice message in the local dialect led to a 3.8 percentage point increase in listening rate (Control mean: 73.2%). Receiving a voice message in the local dialect led to a 7.0 percentage point increase in the likelihood of listening to the full voice message (Control mean: 48.5%).