By Mohamed Ali - Technology & Digital Economy Commentary
Somalia, October 19, 2026 - Get-Phone, a Kenyan company that finances smartphones, has completed its first quarter in Somalia after partnering with Hormuud Telecom. The early numbers offer a quiet but useful lesson for fintech builders across Africa.
The company began in Kenya, later expanded to Tanzania, and in May 2026 signed a partnership with Hormuud. By 1 October it had sold more than 7,000 phones, mostly through 12 Hormuud branches in Mogadishu. Eighty-two per cent of the buyers had never owned a smartphone before.
The package was simple: a phone, 2GB of data a day and 30 minutes of calls for 85 US cents daily. Buying the same 2GB of data alone would have cost about $1.20.
To decide who qualified, Get-Phone used a credit score built on two things: how long a person had been on Hormuud’s network and how much money they moved through EVC mobile money. Scores under 500 were treated as higher risk. For the first quarter, however, the company gave phones to everyone regardless of score. The goal was to test whether the model actually predicted repayment.
The results were almost flat. Customers scoring under 500 repaid 89.7 per cent of what was due. Those scoring 650 or higher repaid 90.5 per cent. The gap was less than one percentage point. When the company looked at household income, the pattern was even more striking. Low-income customers repaid at 94 per cent, while middle-income customers repaid at about 89 per cent.
Related articles
CEO Omar Abdi noted that fewer than 10 per cent of Somali citizens hold a national ID, so a phone number often becomes a person’s main identity. The company is now building a shared defaulter database with banks. That tool could be useful, but it also raises a practical question: who corrects a wrong record once it spreads?
The early data from Mogadishu suggest that willingness to pay and a credit score are not the same thing. A model can rank people, but real behaviour still needs to be measured. Get-Phone’s approach,giving the product to a wide group first, then checking the numbers, offered a clearer picture than the score alone.
For founders building credit products in markets where formal data is thin, the lesson is straightforward. Test the model against actual repayment. Budget for being wrong. Use data and AI, but keep a human who can overrule the machine when the numbers and the people tell different stories.