Monday, September 28, 2009

Some thoughts on the liquidity at agents

One of the biggest challenges in rolling out mobile financial services is the establishment and the effectiveness of the agent network. Agents are the touch-points where the subscribers of the service can get money into and out of the system. (Agents are often also referred to as cash-in and cash-out points).

In instances where a subscriber arrives at an agent with the need to withdraw a large amount it does happen that the agent do not have enough cash to satisfy the cash-out request. This leads to frustration and is one of the reasons why take-up of these systems are slower than what is expected. This problem is referred to as the agent liquidity problem - how to ensure that the agent has sufficient cash available to satisfy the need of the system.

This problem is often approached in a way where the system keeps track of the actual cash available in the drawer of each agent in order to guide subscribers where they can withdraw big amounts. This approach is overtly complex and often fail because of the informal nature of agents businesses. I have recently seen a system that works on averages (what is the average pay-out at an agent or in a town etc.). The system also calculate deviations from real-time on the basis of the amount. For instance, if you want to withdraw $100 you will be able to do so in six hours. This enables the agent to collect sufficient cash if they are pre-warned about a pending withdrawal.

It is plans and innovations like this, that will help deliver financial services to the poor.

2 comments:

Unknown said...

I do agree with you. I think that it will be very tough to get a big amount at an agent even if this business is formal. The mobile banking is intended to small payment.
Thks.

Tim Mukata said...

I am currently designing an MMT solution for a large African Operator.

To address the float problem, I tied a locate agent function to the amount in the till / float, and linked to a customer's desired transaction amount.

This is the approach you have mentioned. You have improved the idea to give it a statistical projectory approach. Fine.

But after some thinking, I shelved the idea because of resulting security risks.

If I want to withdraw $500, the system tells me or a prospective robber which agent has it. Done repeatedly over some days, it will give an idea on the kind of volumes an agent moves.

Too risky in some countries / communities...at least in my opinion.