Mobenzi: Creating Jobs with Mobile Tasks

Posted by AnneryanHeatwole on Jan 31, 2010

Update October 2011:

Last year, we covered the pilot test of the Mobenzi Mobile Researcher, a mobile tool used to collect data and parse text responses into aggregate-able data. Now the tool has been renamed Mobenzi Researcher and is being used in data collection projects in South Africa, Nigeria, and Zimbabwe.

Mobenzi Researcher is a mobile application for collecting data and creating surveys. The app is available in JAVA, Windows Mobile, and BlackBerry versions; the JAVA application is able to work on most feature phones, while higher-end handsets running Windows Mobile or BlackBerry OSs can use extra features like location mapping and image capture. The application works as a series of forms; users collect data by answering a series of pre-determined questions; the data is stored on the phone until it can be uploaded when the phone has network connectivity. This way data isn't lost when users are recording information outside areas of network connectivity. Users can either respond to multiple choice questions, or submit responses with text and numbers.

In The Field

The Mobenzi Researcher has been deployed in several areas now; in Nigeria, The Kano Conditional Cash Transfer program used the survey application to conduct baseline research on school attendance rates of young girls in Nigeria's Kano state. The surveys were deployed across 300 different schools in the state, targeting 13,000 potential students. In Zimbabwe, Africa AHEAD surveyed roughly 1000 households that were members of a local Community Health Club to capture data on a variety of health indicators just as household demographics, health, and hygiene. And in South Africa, a program from the South African Medial Research Council studying neonatal care and prevention of mother-to-child HIV transmission used Mobenzi Researcher to collect data on 25,000 households in the Kwazulu-Natal region. Community health workers used the program to fill in surveys based on their beneficiaries' responses to questions about neonatal care and to aggregate the data in a central database.

Original Post January 2010:

In South Africa, national unemployment stands at 24% – and that number is even higher in many areas outside of major cities. Mobenzi, a new tool created by South African company Clyral, is a project to target unemployed workers to allow them to work over their mobile phones. “Mobenzi agents” use mobile phones to complete tasks that are difficult for computers to process.  

Mark Fowles, co-founder of Clyral, said in an interview with MobileActive.org that the pre-pilot run of Mobenzi at the end of 2009 went better than expected. The original plan was to start a formal pilot in April 2010, but the company decided to run a pre-pilot in order to test the feasibility of the program and to iron out any problems before the main run. 

The initial two-week run focused on text-to-form tasks – taking unstructured data and compiling and synthesizing it into reportable data. The program concentrated on two main areas: sentiment analysis and improving SMS-to-computer compatibility

Job Task 1: What is the Sentiment behind a Message?

The first tasks involved Mobenzi agents reading SMS and Twitter feeds relating to certain topics, and then deducing the sentiment behind the message. One part of the pre-pilot involved following Twitter feeds about local restaurants and interpreting the tweets in a way that a computer can’t; for example “Something about [restaurant] always makes it seem like a good idea…It never is… NEVER”.

Although the sentiment in this tweet is negative, a computer might have trouble parsing that information since the beginning of the tweet is positive. This is where the company’s human intelligence comes in; Mobenzi agents receive a stream of texts and tweets about a topic and then follow a checklist on their phone that categorizes the information; i.e., is the message positive or negative and if it is negative, noting what was the problem, and how the problem can be resolved.

Mobenzi agents act, in essence, as a translation service for data, interpreting meanings and parsing text speak into concrete information. 

Job Task 2: Converting Text-Speak to Language a Computer Can Parse

The second task used Mobenzi agents to convert text speak into computer-compatible forms. Agents would receive a text [for example this from the Mobenzi blog: “im looking 4 brick laying work in Dbn next week. John”], and then transform the information into a computer-ready format [example translation of previous text: “Name: John. City: Durban. Job: Bricklayer. Available: 23-11-2009”]. 

Here, Mobenzi focused on SMS and specifically the usability of SMS systems during competitions, surveys and registrations. These programs usually require users to enter information in a fixed format in SMS (i.e., name, age, birthday). Fowles thinks that allowing people to text a computer service in text speak that is then translated/interpreted by a real person (in this case a Mobenzi agent) will make it easier for people to respond and interact intuitively with services via SMS.

Fowles explains the process: “…Using real people as an interface between the public on their cell phones using SMS and a computer system on the other side. We take free text on one side and we use the Mobenzi agents to structure what once was free text into something that a computer can understand and process.”

How Did Mobenzi Fare?

Mobenzi was built to run on low-end, affordable devices. “We are quite confident that there are tens of thousands of people in our local community who would have compatible phones,” said Fowles, when asked who could operate the Mobenzi platform. Mobenzi agents run the program on their own phones, so high usability was a key factor in designing the tool. “One of the main factors of what we’re trying to do here is these agents – we call them “Mobenzi agents” – we might never even meet them. They might be introduced from another agent, and they can work from their own home with their own phone and never even come into our office. […] Installing the application will be very simple, it will just take a few minutes, and all the instructions for the task are built into the tasks themselves so there’s no training required other than perhaps an hour or so with someone who might be a Mobenzi agent themselves,” Fowles added.

According to Fowles, the initial pre-pilot was successful. “We were very happy with how quickly the guys picked up the concept. Really, on the first day when I went down there with guys I’d never met before, they’d never heard of [Clyral’s other program] Mobile Researcher or Mobenzi. And within an hour the guys were completing tasks. The following week we had an extra 20 people join the pilot and I really hardly did anything – the five guys that were there the week before very easily introduced the concept.”

The goal is now to roll out the next phase of the program in the next month or so, starting with a few key members and then slowly scaling it up. Mobenzi is formally slated to launch in April. According to Fowles, “ we are hoping that it will actually turn into a commercial service instead of running a pilot and ending it – that would be ideal. […] once we’ve got 100 people, or 50 people completing tasks, and we find a specific business client that’s interested in the service, then we might pilot it with them and not charge them for the service. But if it works, we might turn it immediately into a commercial, ongoing business and from that point onwards it would not be a pilot.”

For now, the venture is funded by the Shared Growth Challenge Fund, a fund that was put together by the business trust in South Africa strictly for commercial entities that support projects that foster growth in the very low income sectors, specifically targeting ‘urban poverty nodes.’ One of the urban poverty nodes is just outside Durban, where Mobenzi is based. The Shared Growth Challenge Fund and Mobenzi are partnered with the Shared Growth Challenge Fund matching Mobenzi’s expenses 50/50. In the pre-pilot, Mobenzi agents were compensated for their time, but in the next phase Mobenzi is working on developing a remuneration system that pays per task. 

The pilot was conducted in English, so the agents all had basic grade 12 educations and were fluent in English. However, Fowles pointed out that the Mobenzi concept could be easily adapted to fit into local languages.

For now, the program is still just ramping up. However, if it is possible to turn into a sustainable social enterprise, Mobenzi could provide employment to an area desperately in need of jobs in an extremely accessible way – all the agents need is a phone. 

Photo credit Mobenzi 

Mobenzi: Creating Jobs with Mobile Tasks data sheet 6792 Views
Countries: South Africa

Post new comment

The content of this field is kept private and will not be shown publicly.
  • Allowed HTML tags: <a> <em> <strong> <cite> <code> <ul> <ol> <li> <dl> <dt> <dd><p><br> <b><i><blockquote>
  • Lines and paragraphs break automatically.

More information about formatting options