Man vs. machine: How to implement user-friendly IVR surveys in Somalia

Interactive voice response (IVR) calls are not exactly a user friendly experience – they are known as ‘robocalls’ for a reason. At worse, an IVR call feel like submitting to an interview with Darth Vader. Making IVR calls a user-friendly experience has been one of our main concerns in Somalia in recent months. If we succeed, we could gain the benefits of automating some of our data collection in the country.

WFP/Mohamed Ali Ahmed

                          Internally Displaced Persons (IDPs), Galkayo area.                                       Credit: WFP/Mohamed Ali Ahmed

In February and March, the WFP Somalia team has experienced some success in introducing IVR, by adapting the tool to local habits. It seems that a few critical elements came together:

  • First, our field team are now completely comfortable using Verboice and the IVR system. While we went through some months of trial and error, at present the technology itself is no longer an obstacle. This is important as this allows the field team to experiment and test different approaches at will.
  • Secondly, Somalia gradually introduced IVR after 7 monthly rounds of voice calls that were placed by operators. In Central Somalia, a group of randomly-selected people have been receiving voice calls from our operators every month since April 2014. During this time, people become more comfortable with the survey with our operators.  People who received IVR calls were therefore ‘experienced’ survey respondents who knew what question was coming.
  • Thirdly, the Somalia team understood that receiving an IVR call was somewhat unsettling for survey respondents, who might merely hang up once the call began playing. In order to avoid this reaction, operators called up respondents in order to warn people that they would shortly be receiving an automated IVR call. Response to IVR calls was much better after being cajoled in this way.
  • Fourthly, our questions are worded in the local Somali dialect, and sound familiar to people hearing them.
  • Finally, the team is continuing to try new ways of improving the IVR experience. For instance, they are attempting to re-record our survey questions with the voice of a popular local radio host – someone whose voice was clear, engaging and familiar.  They plan on using these audio files in May.

The use of these different tactics has been met with modest success, with 40 or so calls in February and over 80 in March. Over 50% of people who start an IVR survey finish it. We still have to analyze the data to understand what demographics are successfully responding to IVR calls.

This seems to confirm an important piece of advice we were given by the Listening to Dar team earlier in the project — introduce the tool progressively with the community. The use of principles relating to human-centered design also seem to have paid off.

2 thoughts on “Man vs. machine: How to implement user-friendly IVR surveys in Somalia

  1. Hats off to the team for getting to a point where IVR is working in Somalia. Very much appreciate your careful consideration of the human interaction here and not making it all about the technology as often happens in ICT4D work. I’m curious if you’ve thought about other use cases for the IVR system. It would be compelling to see how Verboice can interact with other systems and be used to deliver content to respondents, not just receive survey responses. For example, getting weather reports into Verboice and allowing a farmer to dial in and get a forecast would be a powerful use-case. Or to have information on food prices at local markets made available through the IVR.

    Keep up the great work!

    • Dear Amit, thanks for your comment.

      You ask what our thoughts are about other use cases for IVR. Our experience to date suggests we could try IVR to collect different types of data, perhaps targeting ‘key informants’ – people in the community that we could train and then call up from time to time using the IVR we have set up. We could think of obtaining information on food and livestock (?) markets with key informants. The system could be used in this way to collect information about health and nutrition too.

      You mention the IVR could be used to make information available to the people in the communities we work in. We are currently figuring out how to do just that in the communities in Somalia and DR Congo where we are now working. First, we are evaluating what priority information needs are. Then, we will make this information available ‘on demand’ at high frequency on the IVR . There will be some technical challenges, including making Verboice work with a shortcode so that calls are free, but we’re confident that we will work something out. Stay tuned for more.

Leave a Reply

Your email address will not be published. Required fields are marked *