Digital Money: Agents are the Key to Mobile Money Transfer

Previous | Next

Summary: New data from M-PESA shows clearly how its success is directly related to its network of trusted agents - a critical success factor for any mobile money project.

New data from M-PESA shows a strong correlation between agent and subscriber numbers. This is further support for our repeatedly argued point that the biggest possible network of trusted agents - shops, street vendors, anyone who sells your pre-paid airtime vouchers - is vital to the success of any mobile money project. Get the agent model - their recruitment and compensation, the wholesale process that supplies them, the OSS back-end that settles the transactions, the fraud prevention and audit measures applied to protect users' money - right, and you're 90% of the way to success. A presentation by Safaricom's head of investor relations, Les Baillie, allowed us to quantify this with actual data - specifically two charts, one of subscriber growth and one of agent recruitment.

We read off the data from the charts, and plotted the two series together. The orange line shows subscriber numbers, on the left scale; the blue line shows agent numbers, on the right scale. The two series appear highly interdependent; the correlation coefficient is 0.96, or near perfect correlation.


To take the analysis further we graphed subscribers against agents using a semi-log plot to cope with the scale difference between millions of subscribers and thousands of agents. We found a strong linear relationship between them, with an R-squared of 0.93, with the slope of the curve - i.e. the marginal productivity of an agent in terms of subscribers - being given by 685x. That is to say, during Safaricom's roll-out of M-PESA, adding an extra agent was worth 685 new M-PESA users.


A possible criticism of this analysis is that the causality might be reversed; more subscribers demand more agents. But it's impossible to become an M-PESA subscriber without access to agents who handle M-PESA transactions, so agents have to lead subscribers to some extent. And history supports this view.

The original roll-out of mobile telephony in the 1980s, and even more so the roll-out of GSM in the 1990s, was powerfully boosted by the ecosystem of resellers and other agents that grew up around the operators, seeking out new markets. 80s veterans in the UK recall resellers who worked out of their car boots, calling regularly at Vodafone's wholesale division to fill the car with handsets.

Later, the launch of pre-paid GSM in the late 90s made mobile a mass proposition and sent us on the way to 4 billion subscribers. And nothing was more critical to making pre-paid work than getting the vouchers out to the market - which meant, crucially, agents. This went double for the roll-out into the emerging markets in the 2000s. It has always been the case in the GSM industry that the macro-economy of network operators and infrastructure vendors has been interdependent with a micro-economy of resellers, agents, shops, MVNOs, and God knows what else.

[Ed. - we'll be discussing Mobile Money business models at the EMEA and America Telco 2.0 events in November and December and sharing new Use Case work with the participants.]