Bank Branches vs. Mobile Money Agents: How Two Infrastructures Serve Financial Access

For most of financial history, the bank branch was the front door to the financial system. To open an account, deposit money, or get a loan, you walked into a building staffed by tellers, managers, and security guards. Branch density — branches per capita — was the standard measure of financial access.

Mobile money changed that. Instead of branches, mobile money relies on agents — local retailers, kiosk operators, and small business owners who handle cash-in and cash-out transactions. In the most agent-dense markets, there is now one financial access point for roughly every half-dozen adults, a density that traditional banking has never approached.

This article compares the two infrastructure models — branch-based and agent-based — across the dimensions of density, cost, reach, and the role each plays in financial inclusion. For the full data ranking of 58 economies by agent density, see our mobile money agent density rankings 2026.

The Scale Difference

The most striking difference between branches and agents is scale. In Nigeria, the density gap is more than 600-to-1:

• Bank branches: 26.6 per 100,000 adults (roughly one branch per 3,700 adults)
• Mobile money agents: 17,665.9 per 100,000 adults (roughly one agent per 6 adults)

That ratio — roughly 664 agents for every bank branch — is the widest in the dataset, but the pattern holds across almost every market where mobile money is established. Even in more moderate cases like Kenya (1,069.2 agents vs. 5 bank branches per 100,000 adults), agents outnumber branches by more than 200 to 1.

The comparison is not apples-to-apples. A bank branch is a full-service facility with professional staff, security systems, and regulatory compliance infrastructure. A mobile money agent is often a corner shop that offers cash-in/cash-out as one service among many. But the density comparison matters because it reveals how each model scales.

Why Agent Networks Scale Faster

Agent networks scale the way retail franchises scale — through existing commercial infrastructure. A mobile money operator does not need to build a branch; it needs to recruit an existing shopkeeper, train them, and equip them with a basic point-of-sale device or phone.

This has three implications:

1. Capital costs are near-zero per agent. The mobile money operator does not pay rent, utilities, or teller salaries for each agent location. The agent absorbs those costs as part of their existing business.

2. Deployment speed is measured in months, not years. A mobile money operator can onboard thousands of agents in a quarter. Building thousands of bank branches would take decades.

3. Density follows demand naturally. Agents appear where there are people and transactions. The network self-organizes around economic activity in a way that branch networks, planned by bank headquarters, cannot match.

The result is that agent density in the top 10 mobile money markets ranges from 3,140 to 17,666 per 100,000 adults. By comparison, the top 10 economies for bank branch density range from 32 to 122 per 100,000 adults — still two orders of magnitude lower.

What Each Model Does Best

Agents and branches are not perfect substitutes. Each infrastructure excels at different functions.

Branches excel at:

• Complex transactions: Loan applications, investment products, account opening with identity verification
• Dispute resolution: A physical office where customers can escalate issues
• Trust and safety: Regulated premises with insurance and security
• High-value services: Wealth management, corporate banking, international transfers requiring documentation

Agents excel at:

• Cash-in/cash-out: The most basic financial function — converting cash to digital money and back
• Geographic reach: Penetrating rural areas where bank branches would be uneconomical
• Transaction frequency: Handling large volumes of small-value transactions
• Extended hours: Many agents operate beyond traditional banking hours

For a deeper comparison of how these models serve different populations, see our analysis of how mobile money bridges financial inclusion.

The Hybrid Model

The most effective financial systems use both infrastructures in combination. In Ghana (4,241.2 agents per 100,000 adults), mobile money agents handle day-to-day cash transactions while bank branches provide credit, savings products, and formal financial services. The two systems are complementary, not competitive.

This hybrid model is visible in the data. Markets with high agent density tend also to have growing account ownership rates. Ghana’s 81.2% account ownership (2024 Findex) puts it in the top third globally, significantly higher than the Sub-Saharan African median. The agent network did not replace the banking system — it extended its reach.

The same pattern appears in Senegal (2,608.6 agents, 76.5% account ownership) and Tanzania (3,744.6 agents, 59.8% account ownership). In both countries, mobile money agents act as a gateway: customers first use agents for basic transfers, then graduate to fuller financial engagement through bank-linked mobile services.

Where Branches Still Matter

Agent networks have limits. They cannot originate loans, provide financial advice, or handle complex regulatory compliance. For these functions, bank branches remain essential.

The countries with the highest bank branch density — San Marino (122 per 100,000 adults), Bulgaria (92), Bolivia (62.6) — are not necessarily more financially inclusive than high-agent-density countries. But they offer a different kind of access: deeper, more formal, and more integrated with the broader financial system.

For businesses, branches are especially important. A firm that needs a bank loan for investment or working capital cannot get it from a mobile money agent. Branch-based credit infrastructure — credit bureaus, collateral registries, loan officers — is irreplaceable for business finance.

The Cost Question

Agent networks operate at a fraction of the cost of branch networks. The IMF estimates that a mobile money transaction processed through an agent costs roughly 30–50% less than the same transaction processed at a bank branch, primarily because the agent model avoids fixed infrastructure costs.

This cost advantage is critical for financial inclusion. In markets where per-capita income is low and transaction sizes are small, bank branch economics do not work. A branch costs the same to operate whether it serves 500 customers or 5,000. An agent, by contrast, costs almost nothing to establish and only generates costs when transactions occur.

That cost structure is why Nigeria can support 17,666 agents per 100,000 adults but only 26.6 branches. The agent model is viable at a transaction volume that would never support a branch.

Conclusion: Two Infrastructures, One Goal

Bank branches and mobile money agents are not competing models — they are complementary layers in a financial system that is becoming more inclusive by using both. Branches provide depth (credit, savings products, complex services). Agents provide breadth (geographic reach, transaction frequency, low-cost access).

For the roughly 1.4 billion adults worldwide who remain unbanked, the agent model is often the more practical path to financial access. For those already in the system, branches provide the services that agents cannot. The question for financial inclusion policy is not which model to choose, but how to build both — connecting agent density to branch depth to create a seamless financial infrastructure.

For the data behind agent density rankings across 58 economies, see our companion mobile money agent density report.

Sources & Method

Mobile money agent density data comes from the IMF Financial Access Survey (indicator FA57N.PHTADLT_NUM — agents per 100,000 adults). Bank branch density data come from the World Bank World Development Indicators (indicator FB.CBK.BRCH.P5 — commercial bank branches per 100,000 adults). Account ownership data are from the World Bank Global Findex Database 2024. See the methodology page for full indicator definitions and dataset construction notes.

Cost comparisons between agent-based and branch-based transaction processing are based on IMF FAS documentation and industry estimates cited in the Findex database methodology.