Author: Marie Mallet

This blog post is part of a series on SMS surveying. The first part, ‘SMS Surveying: Design’, can be accessed here. The third part, ‘SMS Surveying: Lessons Learnt in Tanzania’, can be accessed here. 

SMS surveys are fast and easy to implement without the need for field logistics and personnel for conducting the survey, apart from the programming and data processing tasks. However, there are still a few important logistical aspects to consider before implementing an SMS survey and these are summarised below following my recent experience in Tanzania.

    • Choosing the type of system for sending SMS between short code1 SMS and Unstructured Supplementary Service Data (USSD). There are two main systems to allow the sending and receiving of text messages from a large number of respondents, namely, by using the short code SMS or using a USSD extension. These two systems can be used for SMS surveys and both allow for reverse billing, which means they automatically charge the messages sent by the respondents to the organisation hosting the SMS survey. The main difference between the two systems is on the storage and display of the messages. The SMS content remains stored in the respondent’s mobile phone memory while the USSD does not. In addition, a USSD question can have up to 182 characters and is more interactive as it reaches respondents slightly faster than SMS by creating a real-time connection (see screenshot on the left in Figure 1 below). However, using USSD for an SMS survey may be risky as there is a time limit to respond and the survey can be easily interrupted if, for example, the respondent receives a phone call or stops the survey by mistake. Also, the local context and the type of population targeted should be considered when choosing between the two systems as some mobile phone users may not be familiar with using USSD compared to using SMS.

Figure 1. Mobile screenshots showing USSD display (left picture) and SMS display (right picture)
    • Short code number license in the case of the short code SMS: To be able to send and receive SMS, the organisation implementing the SMS survey needs to access a short code number from which the respondents will receive the message and to where they will text back the answers. In Tanzania, this short code number can be rented using already existing short codes from specialised telecommunication companies or a dedicated two-way SMS short-code can be directly procured from the Tanzania Communications Regulatory Authority (TCRA). For the latter, this process needs to be initiated in advance as it can take up to 10 weeks to obtain.

       

    • Network integration to the short code number: When acquiring a dedicated short code number, the time necessary for the networks to be integrated within this short code number needs to be taken into account. The integration of each available network in the country to the two-way SMS short code enables the researcher to contact and receive SMS to/from the respondents on these networks. In Tanzania, this process can take time as there are four networks (Vodacom, Tigo, Viettel and Airtel), each with a different structure and resources, which affect the time to integrate the network to the short code. It is therefore advised to initiate this process at the earliest possible date to avoid delays in survey implementation.

 

  • Adding a round of phone calls before launching the SMS survey: Depending on the research objectives, the number of SMS survey rounds, and sample size, the researcher could consider having a first round of phone calls before launching the SMS survey. The purpose of these calls is to introduce the SMS survey to the respondent, explain how and when it will be implemented, how the system works and the fees charged for the reply SMS sent by the respondents. Adding a round of phone calls may be preferable in the case of a follow-up survey where limiting the attrition is important and where replacement/increasing the sample size is not possible in case of non-response. Calling the respondent in advance would help maximise the answer rate, and check eligibility of the respondent before the SMS survey data collection, e.g., to confirm the phone number, to check the respondent’s ability to read/send SMS, etc. This round of phone calls could also be an additional opportunity for the researcher to collect data over the phone by asking additional questions. However, adding a phone survey round has a cost and may not be suitable for SMS surveys with a large sample and where a low response rate can be addressed by increasing the sample size.

     

  • Respondent incentives and SMS cost: Depending on the context of the survey, providing compensation to the respondent after completion of the SMS survey to thank them may help in keeping them engaged, especially for multiple SMS rounds surveys. The incentives can take the form of a small token provided by mobile money, however, there is a cost to be considered for sending/receiving the transfer that differs across the networks.

     

  • Timeframe for SMS data collection: A one-week window is enough for covering the sample (regardless of the size) and when using reminder messages. From our experience in Tanzania, with a project that consisted of four survey rounds with a one-week timeframe; among the surveyed sample2 75%3 of respondents completed the survey in the first 24 hours after the launch of the SMS survey, and 95%3 within the first two days. In each SMS survey round, the number of additional respondents who answered on the day 3-7 period was limited (<5% of the total surveyed sample), despite the daily reminders sent to the remaining sample. These results suggest that the answer rate obtained in the first days of data collection is likely to be similar to the final answer rate at the end of the SMS survey and extending the SMS survey by more days would have a marginal impact on the answer rate.

This blog is part of a series on SMS Surveying. In the final blog of this series, I will share the lessons learnt and recommendations on SMS surveys from my experience of conducting an SMS survey in Tanzania in late 2020.

 

1A SMS short code is an abbreviated phone number and consists of 5 or 6 digits in length.
2Sample of respondents who started or completed fully the SMS survey.
3The percentages were computed from the total of the interviewed sample and exclude the respondents who have not answered the survey, therefore these percentages do not reflect the actual answer rates that were varying across the rounds and ranging from 31% to 72%.