Everything You Need to Know About USSD vs SMS

After we published our software engineer, Alex Muller’s, blog on USSD, we realized there was more to say about how USSD compares to SMS.

First, a run down on costs. USSD cost depends on the amount of time that it takes people to go through the questions as well as how many sessions it takes. It’s a little complex so we can break it down as follows:

Aggregator fees: Each time a person dials into the line, they start a new session. A session is measured in increments of 20 seconds and last up to a total of 180s before the session is terminated. Aggregators charge a fee for the initiation of each session. Therefore if someone dials in for 35 seconds the organisation will be charged the same fee as if they dialed in for 19 seconds. This is the fee that an aggregator will add to any network charges. If the organisation doesn’t use an aggregator, they will not be charged this but then there will be maintenance costs for running the integrations with networks.

Network Fees: Network fees are based on the amount of time that people spend on the line and will depend on the network and what rates you negotiate with them but as a general rule it costs around ZAR0.2/ 20 seconds for USSD. It very much depends on the MNO (mobile network operator) and if you have negotiated a rate based on volume with them. . If the service is zero-rated, the amount paid by the programme doubles as you are paying for the user’s ZAR0.2/20 seconds cost too.

In summary, we would generally say that for a five or six question USSD survey, it would cost around $0.3. Generally it doesn’t matter what network the user is on, although in some countries some Network Operators are not yet offering USSD as a commercial service. This doesn’t mean that they won’t have USSD capabilities as all operators have this as a base service, but they have to agree to offer this commercially.

Timed Sessions: USSD sessions are ‘conversations’ with a computer and as such are constrained to a certain time limit. Time is usually limited to between 90 and 180 seconds. This limit has to be negotiated with the WASP. Due to this time constraint sessions should be aimed at getting information from the user as efficiently as possible. While it’s true there is no time-out for SMS, this can also serve as a challenge as the user is not incentivized to respond in a certain amount of time.

In our experience, a five question multiple choice questionnaire can generally be completed within 180 seconds. This is ideal because one of the tricky things about USSD is that it terminates at 180sec whether you have completed the questions or not. We mitigate this by ensuring that if people dial back into a line they are inserted in the flow where they left off. But if it takes many more than two 180sec sessions to complete a questionnaire there will likely be a major drop off.

Character Limit: There are only 140 characters per USSD page allowed which translates to approximately one question and 6 options for the user to choose from. The SMS character limit is 160.

Because of these factors, while we have found that USSD is much better for gathering different pieces of data than SMS (structured or unstructured) there are some best practices to follow:

Question limit: Questions should be limited to a maximum of six questions per session. This includes only one free form text input. The number of questions decreases as more text inputs are added. There is no question limit for SMSs.

Inputting Information: As much as possible users should be given a choice of responses to select rather than inputting text. This is good practice for both USSD and SMS. Generally users in most USSD campaigns have alphanumeric keypads and inputting text can be erroneous and time consuming.

Inputting information for SMS is more complicated as there is a back and forth which can increase drop-off rates. Inputting information for SMS is not necessarily more complicated than USSD — they can be identical. For USSD, drop-offs are often caused by time-outs. For SMS, drop-offs may be caused by network latency (i.e. the MNO taking a long time to deliver SMSes). In our experience SMS delivery can be very fast (a user would receive their response SMS within seconds).

User Details: Capturing user information is essential but users don’t want to be bombarded with questions about themselves. If the campaign is ongoing users’ information should rather be built up piece by piece.

In conclusion, a quick overview of benefits of SMS include familiarity, accessibility on most phones, and they can be pushed and remain on a person’s phone. USSD allows the user to choose and navigate from multiple options, is also accessible and familiar, but is limited because of timeouts and may lack commercial viability in some countries. During recent user testing on our mHealth platforms in South Africa, users found a survey over SMS far easier and better UX than over USSD. But the costs of an SMS surveys are higher. So there are pros and cons to both.

Contributed by Lauren Kotze, Strategic Lead, Praekelt.org