Several years ago, regulation was the apparent barrier everyone spoke about in branchless banking. Today, it seems like everyone is on about agents. With good cause. Agents are the customer-facing element for providers, who rely on them to open accounts, do customer care, and (crucially) stock adequate amounts of cash and e-float to enable clients to deposit and withdrawal.  Yet, there is no consensus on how to build a viable agent network.

CGAP looked at M-PESA merchants in Kenya for clues about the profit drivers for agents . We studied 20 agents with 125 locations. We focused on small stores of the kind found in urban slums and rural areas, which make up the vast bulk of M-PESA agents. We spent 3 weeks in the field. What did we find?

First, it is profitable, to the tune of 3.2 times more profit per day being an agent (US$ 5.01/day) than selling airtime (US$ 1.55/day). Quite impressive.  What are the drivers behind this? First, volume of transactions. On average the agents we looked at did 86 transactions per day. But it’s equally critical to understand costs. In fact, this is where providers will need to do the most work to understand their agents.

The number one cost for most agents was liquidity management – moving cash. Agents report a host of expenses, including bank charges, transport costs, and fees to aggregators who advance commissions and provide easy float/cash swaps for agents. On average, liquidity management consumed 30% of total expenses. Extending the network of aggregators would help alleviate some of the costs, and Safaricom has taken steps to calibrate the fees aggregators charge.

Second, being an M-PESA agent requires capital, and that capital has a cost. The typical agent had US$ 1,605 tied up in the agent business, mostly in cash and e-float. This is 12 times greater than the US$129 capital they had invested in airtime (mostly in their stock of scratch-off cards). To put this in further perspective, US$1,605 is equivalent to Kenya’s per capita GDP.  Many agents we spoke to financed their initial capital for the agent business out of proceeds from another business. But the smaller the agent, the more frequently they reported borrowing, often in sums which are large compared to their other income flows. There are also a good number of M-PESA shops which were started from scratch, and these are usually financed by a loan. Many of these entrepreneurs are barely breaking even. Cheaper financing, secured by their M-PESA commissions, would help them immensely.

Overall, M-PESA’s wild success – with 1 in 4 adult Kenyans signed up in 30 months – has smoothed out a number of potential bumps in building a large agent network. But if Safaricom had had more modest success , it could have found difficulty in enticing merchants to be agents. Our research shows that, at the current commission rates and expenses agents reported, most agents would be unprofitable under 30 transactions/day. Providers in other countries should take note: few of their agents average this many transactions. In other words, M-PESA’s commission structure wouldn’t work elsewhere. Providers will need to do the grunt work of calibrating a commission structure to local circumstances.

You can see more details here. Some of these findings were presented at the Mobile Money Summit in Barcelona. The fieldwork was done by myself and Sarah Rotman, Ignacio Mas, and Olga Morawczynski (whom many of you will recognize for her ethnographic research on M-PESA customers).

-Mark Pickens