With guidance from Text to Change and in partnership with local media organizations, TRAC FM provides citizens in Uganda with a platform to monitor, scrutinize, and discuss public service issues via SMS, radio, and online data visualizations. It does this via the TRAC FM software, a tool for Ugandan radio stations that is built on an open source platform.
A recent radio poll at Sanyu FM in Kampala, Uganda, asked listeners what area of service delivery should be a priority: healthcare, education, security, sanitation, or transport. Using a new tool, TRAC FM, the station was able to solicit comments via SMS from listeners, discuss the issue on the air, and create and post online visualizations of the responses. The station received 103 SMS responses which showed that healthcare was the major concern for listeners, which accounted for 65% of responses.
With help from Text to Change and in partnership with local media organizations, TRAC FM provides citizens in Uganda with a platform to monitor, scrutinize, and discuss public service issues via SMS, radio, and online data visualizations. It does this via the TRAC FM software, a tool for Ugandan radio stations that is partially built on RapidSMS, an open source platform originally developed by UNICEF.
According to the TRAC FM website:
TRAC works in countries affected by poverty and conflict. Its sole purpose is to improve the welfare of people by enabling them to make informed choices and to hold their leaders accountable. TRAC gathers data to enhance transparency and informs people by unleashing the power of mobile communications.
The TRAC FM software platform was tested earlier this year with students from the Makarere University Business School in Kampala. Now,it’s being used at two radio stations in Uganda. MobileActive.org spoke with Wouter Dijkstra, program manager at TRAC, to hear more about how it works.
Where TRAC FM is Being Used
Dijkstra and his team launched the first programs in July at two radio stations: Radio Wa in Lira in the north of Uganda and Sanyu FM in Kampala, the city’s oldest radio station.
Rather than fit the station to the software, the TRAC FM software is embedded into existing radio programming.
At one station, the software is used to report on failed service delivery with a weekly talkshow featuring prominent guests who comment on both the topic and the collected data. During the rainy season in Uganda, lightning strikes are a major concern. Dijkstra said that during a recent season, 40 school pupils died because of lightning strikes. At Radio Wa, people in Lango were asked to report on the presence of lightning conductors on school buildings, a requirement for every school. The results were discussed with the regional education commissioner during the show.
Another example is the opinion poll that was posed to listeners of Sanyu FM in Kampala, in which 65% of respondents said that healthcare was the most important service delivery issue. Data visualizations are then created from poll results.
For more, see also this video which explains more about TRAC FM.
How TRAC FM Recognizes Text
TRAC FM software recognizes what people “want to say” in three ways, Dijkstra said.
- A string distance algorithm is used to help filter out minor typos. If an SMS with a spelling error comes into the system, it compares the word to all possible answers given for the poll. For example, if a listener sends in “healthcare” instead of “healthcare” the system sorts the response under “healthcare.”
- Another way the system recognizes keywords is via aliases. For every category in the poll, the system enables rules which link to a certain category. For instance, if a listener responds with the word “hospitals”, and this is an alias of “healthcare,” the TRAC FM system sorts the SMS under the category of “healthcare.”
- If after these two processes the reply is still unknown, the TRAC FM system sorts the word under a category of “unknown” and the team can manually assign the response to the correct category.
The TRAC FM software also plots SMS responses in a data visualization, including bar graphs, time graphs, and maps. Through this, radio presenters can get an overview of where reports are coming from and what the issues are. These overviews also help get feedback to the listeners. “Instead of people just calling in and giving their opinion, you get an entire audience’s response to a certain issue,” Dijkstra said. “This gives it more substance. It’s more valuable data for discussion.”
Radio stations generally do not have websites in Uganda. What they do have is an Internet connection to download audio files they use for advertisements from companies like MTN. So, visualizations of poll results are presented in the station through the online TRAC FM software interface. Because few in the rural areas have access to Internet, TRAC FM focuses specifically on people who only have access to radio and feature phone. Radio, then, is the key delivery mechanism for sharing poll results.
“People without Internet, with only radio and SMS, add contributions through SMS, and get results through radio,” Dijkstra said.
Radio presenters log in to the TRAC FM system through their Internet connection in the studio and share the poll results with listeners during their talk show. This “instant feedback loop using radio, mobile, and Internet is a configuration that is fully adapted to available technologies in rural Uganda,” he said. View more data visualizations on the TRAC FM Facebook page.
Engagement with Listeners
Every listener who sends in an SMS receives a confirmation message. They are also given an option to respond via keyword to join an active user group, meaning that they will receive SMS updates on upcoming polls. In this way, “we are building up a group of respondents who are actively involved,” Dijkstra said. “That is how we are building up a mobile active group.”
Note: For more on using mobiles to engage radio listeners and connect journalists and citizens, see also this case study on how WNYC and radio program The Takeawy established sourcing through texting.
Similar to WNYC and The Takeaway, Dijkstra uses the TRAC FM software to collect additional demographic information to get to know the participants. Through a call center, the team contacts the respondents to ask additional questions like age and occupation. In this way, they are poised to target specific people with specific questions.
The software can also be integrated with social media sites like Facebook. Results from listeners are instantly relayed to the radio station and live updates are given during the show. “In this way, we instantly reach all the listeners,” Dijkstra said. The Kampala radio station has 5000 followers on Facebook, and poll results are updated on the Facebook page so people can discuss the poll online.
And people are using it. During a recent one-hour segment in Lira on the topic of the Regional District Commissioner, the radio station received 269 replies. Hear from program managers in this YouTube video about TRAC FM.
“This form of interactive radio and debate in very new to these regions and gives people a feeling of being important and being part of the solution to the many problems of public service delivery,” Dijkstra said. “This is exactly our goal.”
Challenge: Streamlining the Process
Dijkstra said that one of the biggest challenges thus far has been streamlining the process with different parties involved in the process of running a poll. This includes Ugandan telecom providers and radio station management. Dealing with both “takes a lot of time and effort,” Dijkstra said. “Getting the software and broadcasts up and running requires patience and a personal approach where the right decision takers are required to give the green light to proceed.”
Leveraging Local Languages
Dijkstra said that the biggest success thus far has been the first broadcast in Lango. In this region, people often do not speak English, so the team ran the poll entirely in the local language, Luo. Within 8 hours, the station received 210 replies in Luo. Additionally, 320 people participated in a poll in Lira, telling the station about the accountability process at their local cooperatives while a local policy maker joined in the studio to comment on poll results. The TRAC FM software recognized, categorized, and visualized the responses instantly so radio presenters could update the larger listening audience on the results. “The ease with which the system could be converted to other languages was even a surprise to us,” Dijkstra said.