Before I started Telerivet, I spent two years as a Peace Corps volunteer in Tanzania, working on local ICT skills in communities where a feature phone was often the only computer anyone owned. What stuck with me was not the technology itself, but how much people did with it once they could talk back instead of just receive.
Years later, I found myself reading a randomized controlled trial published in the Journal of Development Economics, run by researchers from the London School of Economics, McGill University, Sokoine University of Agriculture, and CIAT. The researchers built ShambaChat, an SMS-based discussion platform, to test something specific: does peer-to-peer learning, proven to work in person, still work when it happens over text message? Because farmers were randomly assigned to one-way or two-way groups, the researchers could isolate the effect of conversation itself, rather than just differences between the farmers who happened to engage more.
They ran it with maize farmers in rural Tanzania over three monthly rounds tied to the agricultural calendar. Every farmer received the same extension content by SMS. Only some could reply to it and see what other farmers were saying. That single design choice, the ability to talk back, was the entire experiment, and it changed what the same message was able to accomplish.
Farmers with access to the peer discussion groups were 18 percentage points more likely to intercrop legumes with maize, and 15 points more likely to make organic compost from farm waste, compared to farmers who received identical content with no way to discuss it. Those are large effects for hard behaviors. Intercropping and composting require coordination and local adaptation that a one-way broadcast cannot supply.
The mechanism is worth sitting with. Farmers used the reply channel to name specific legume varieties, point each other to nearby seed sellers, and share planting timelines drawn from their own conditions. The extension content set the agenda. The conversation made it actionable. And the effect traveled: farmers in the same villages who were never part of a chat group still picked up the practices, because the information moved through the community once it had somewhere to go.
The study also found something less comfortable. Engagement dropped sharply after the first month, and the adoption effect faded along with it. Reach and sustained attention turn out to be separate problems, even when the reach is interactive. That is a design challenge to plan for, and the approach survives it.
I keep coming back to this study because the farming is almost incidental. The result is about what happens the moment a message stops being a broadcast and becomes part of a conversation.
We see the same pattern across every field where Telerivet operates, from payment confirmations to beneficiary registration to field research programs collecting data across multiple countries. A message that only goes out is an announcement. A message that can come back is the start of a workflow. A platform built only for outbound messaging quietly caps what a program can achieve, whether the goal is behavior change, data quality, or trust.
This is close to what we mean when we talk about communication orchestration: designing for what happens when someone replies, and building that reply into the next step of the workflow rather than treating it as an exception. If that idea is new to you, our practical definition of communication orchestration lays it out in full.
For research teams specifically, this study is a useful data point in an argument I think is underappreciated: an SMS instrument that only pushes surveys or reminders out is leaving information on the table that a two-way instrument would capture naturally through normal conversation, in the form of clarifying questions, local context, and the kind of peer correction that a single enumerator would never catch. We have written before about the operational side of this, in running multi-country SMS data collection for field research, and the mechanics of building reply-driven programs in what happens when someone replies.
None of this works if the channel cannot reach the participant in the first place. ShambaChat ran on basic mobile phones over SMS because that is what reaches rural Tanzania reliably. That is the same principle behind connectivity completeness as we think about it: build for the device and network conditions actually in the field.
Design for the drop-off from day one. The ShambaChat researchers were candid that engagement fell off after the first month, and that the adoption effect faded with it. If your program depends on sustained dialogue, plan for how you will keep the conversation worth returning to, and treat that as seriously as how you will start it.
Give people something real to reply to. The messages that did the work in this study were concrete and specific enough that a farmer could answer with a detail from their own field. Vague broadcasts get silence. Specific ones get replies.
And treat the two-way channel as the instrument itself. The entire effect in this study came from a single design decision: letting people talk back. Everything else about the program was identical.
We were glad to be the platform underneath ShambaChat. It is exactly the kind of program this company was built for, and studies like this one are the clearest evidence I have seen that the difference between reach and dialogue is the whole result.
If you are running or scoping a field research program and want to talk through what a two-way SMS instrument would take to build, get in touch with our team.