No Single Recipe When Structuring Agent Networks

22 February 2011

This is the fourth piece in the five-part series launching CGAP’s Agent Network Management Toolkit (available to download and highlights are on CGAP’s website). The toolkit is based on more than a year of research that yielded data on more than 16,000 agents in Brazil, India, and Kenya. In-depth interviews were conducted with 466 agents, agent network managers and providers, including mobile network operators, banks, MFIs and technology companies.

When providers are planning the launch of a branchless banking service, one of the critical decisions facing them is how to structure the agent network. As CGAP’s newly released Agent Management Toolkit describes, there are many different ways to do this. For example, half of Orange Mali’s agents were MFI branches when it first launched. EKO in India planned to use the huge distribution network offered by the largest telco in India, Airtel. And others like WING Money in Cambodia painstakingly sign up local merchants one by one to act as agents. There is no best structure for an agent network and the toolkit describes some of the implications of different network structures on factors such as network readiness (speed to go to market), reach (number and location of outlets) and control that the provider has.

Yet most agent networks that have been around for a while have one thing in common: their structure has changed dramatically over time. M-PESA in Kenya, which has seen its agent network grow from zero to over 23,000 in less than four years, has gone through three different phases in regards to agent management.

When M-PESA first launched in March 2007, Safaricom selected 400 of its largest airtime distributors to act as agents. In the first months after launch, it signed up several hundred more agents, most of whom were large Safaricom airtime dealers. Although Safaricom could carefully control its dealers, the growth rate of customers was far outstripping the growth rate of agents and Safaricom realized it needed to loosen its controls.

Less than a year after launch, Safaricom realized that growing numbers of ‘aggregators’ were signing up subagents who could not meet Safaricom’s agent criteria to operate under the license of the ‘aggregator.’ M-PESA benefited from these arrangements and allowed them to occur with tacit approval. The agent network grew exponentially to more than 10,000 points. However, the fact that these relationships were negotiated privately meant that much of the agent network was removed from Safaricom’s direct control. This led to problems around adequate float, look and feel of the location.

In late 2009, Safaricom embarked on its third phase of agent management. It reasserted more control over its agent network and announced that each subagent must have a direct contractual relationship with Safaricom. Although this means more work for Safaricom, M-PESA knows it must monitor agents to ensure a consistent customer experience. The number of M-PESA agents has continued to grow but M-PESA monitors agents carefully and may enforce stricter measures to eliminate those that do not meet standards.

What’s next? This latest phase is almost certainly not the last and as the M-PESA service evolves, it continues to face changing pressures that impacts its agent structure. This story shows us that providers will need to maintain flexibility to ensure agent networks grow in line with customer acquisition while maintaining a viable business case for agents.


- Claudia McKay


Add new comment