📱 Mobile money in Africa

What mobile money really is

Mobile money is an e-money account opened with a licensed issuer, usually the financial subsidiary of a mobile network operator. That sets it apart both from a wallet linked to a bank account and from a simple payment app. The account is identified by a phone number, funded and cashed out through an agent network, and operated from any phone via USSD. These four building blocks make up a complete payment system. It relies on no retail banking infrastructure, and that independence explains why it took off in areas the bank branch network never reached.

Global flows have grown sharply in recent years. It took mobile money two decades to pass $1 trillion in annual flows worldwide, and only four years to double that (GSMA, SOTIR 2026). Flow measures how intensively existing users rely on the service. The ratio of registered to active accounts measures how many people actually use it, and that second metric is what drives market sizing. About 1.2 billion accounts are registered in sub-Saharan Africa and North Africa. Only 347 million of them record a transaction in any given month. Roughly seven in 10 accounts therefore show no monthly activity.

Retail payment rails differ so much from one African country to the next that no single collection method covers the continent. Collecting payments in Nairobi means M-PESA, which accounts for about 89% of mobile money in Kenya. Collecting in Lagos means an instant NIP transfer from a bank or fintech account. That is two integrations, two risk models, and two cost structures. Mapping, country by country, the rail that actually carries retail payments is therefore the first step for any entry into Africa, and it is the step most often skipped.

The major networks and their actual numbers

Four mobile money ecosystems cover most of the continent: M-PESA (Safaricom / Vodacom, since 2007), MTN Mobile Money, or MoMo (MTN Group, since 2009), Airtel Money (Airtel Africa, since 2011), and Orange Money (Orange Middle East and Africa, since 2008). Dominant national players add to these, including telebirr in Ethiopia, EcoCash in Zimbabwe, and MVola in Madagascar. A low-cost operator, Wave Mobile Money, later drove down transfer prices in francophone West Africa.

Value is spread very unevenly across the continent’s subregions. East Africa accounted for $806 billion in mobile money flows in 2025 and West Africa for $498 billion (GSMA, SOTIR 2026). Central Africa, with its 20 live services, remains an order of magnitude smaller. Its interoperability was therefore imposed by the central bank rather than negotiated by the market (see GIMACPAY, below).

USSD: the channel that actually carries the market

USSD (Unstructured Supplementary Service Data) is a GSM signaling channel repurposed as a payment interface. The user dials a short code, such as *334# for M-PESA in Kenya or *170# for MTN MoMo in Ghana. A synchronous session then opens between the handset and the operator’s server, and a text menu appears. The channel works on a $15 phone, with no mobile data, no app, and no updates. The addressable base is therefore every GSM handset in circulation, not just smartphones. That reach, not the quality of the interface, is what drove the spread of mobile money.

*XXX#                          operator short code, typed on the keypad
  |
  1. Send money                 -> P2P on-net and off-net
  2. Withdraw money             -> generates a withdrawal code for the agent
  3. Pay a merchant             -> enter the merchant number (till / merchant code)
  4. Pay a bill                 -> biller code + customer reference
  5. Buy airtime
  6. My account                 -> balance, statement, PIN change
  7. Loans and savings          -> short-term credit, overdraft, nano-savings

  Typical user input (branch 1):
    1                            selects "Send money"
    07XXXXXXXX                   recipient's MSISDN
    2500                         amount
    ****                         PIN (mobile money PIN, 4-5 digits)
  Response: confirmation SMS carrying the transaction ID
            -> THIS ID IS THE PROOF OF RECORD IN A DISPUTE

  Session constraints to know:
    - synchronous session, typically times out after 20 to 180 s
      depending on the operator -> any slow external call kills the transaction
    - small screen (~182 characters per USSD screen)
    - no state on the handset: the server holds all the context
    - the channel is not end-to-end encrypted: the PIN protects
      the transaction, not the confidentiality of the transport

• The short code is a scarce, regulated asset. The national telecom regulator assigns it, not the mobile money operator. Getting your own code (as an aggregator, a bank, or a biller) means filing an application with the country’s electronic communications authority, and lead times run to months.
• Codes differ by country, even for the same operator. MTN MoMo is reached via *170# in Ghana, *165# in Uganda, *133# in Côte d’Ivoire, *126# in Cameroon, *182# in Rwanda, and *671# for MoMo PSB in Nigeria (codes published by the operators in 2026. Always double-check with the local subsidiary before integrating: they change).
• The latency budget is the real technical contract. A USSD session times out. If the merchant’s back end takes several seconds to respond, the user sees an error screen and the transaction is left in an indeterminate state. Every USSD integration must be built as an immediate response plus asynchronous confirmation, never as a blocking call.
• SIM Toolkit is the layer above. An STK push (a prompt pushed to the handset asking the user to enter their PIN) is not USSD: it relies on the applet embedded in the SIM. It powers online payments in Kenya (Lipa na M-PESA Online), and it spares the customer from having to remember any number.
• The channel is expensive for the merchant. Operators bill USSD sessions, often per session or per screen. A poorly designed flow that adds screens adds cost, and drop-off.

The agent network and float: the invisible infrastructure

An agent is a merchant appointed by an issuer to convert cash into e-money and back. This ordinary shop (a kiosk, telecom store, pharmacy, or gas station) holds two stocks at once: e-money (its float, or e-value) and cash. A customer deposit drains the float and fills the cash drawer. A withdrawal does the reverse. An agent’s day revolves around managing the liquidity balance between the two stocks more than around selling anything. That balancing act is the No. 1 point of failure in the user experience across the continent.

• The agent’s commission is a distribution cost, not an issuing margin. It comes out of the fees charged to the customer and is passed on to the agent in tiers by amount. When an operator slashes its prices (Wave), it automatically squeezes agent pay and has to make up for it with volume. Otherwise the network disengages.
• Agent exclusivity has disappeared almost everywhere. Most regulators now ban it, so the same kiosk serves M-PESA, Airtel Money, and a bank. That is a win for customers and a constant fight over float availability for operators.
• Agent fraud is a category of its own: deposits not credited, overcharging relative to the official price list, withheld cash, and “helpful” assistance entering the PIN of a customer with low literacy. Operators respond with mystery shopping, mandatory price displays, and cutting off agents at fault.
• Agent float is a working capital need. When an agent closes at the end of the day for lack of liquidity, a customer cannot withdraw their pay. The best merchant rollouts on the continent factor the local liquidity calendar (paydays, market days, holidays) into their collection forecasts.

The e-money framework: licenses, safeguarding, KYC

Mobile money falls under e-money issuance, a regime separate from banking: the issuer extends no credit and does no maturity transformation on the funds it receives. The backing principle is the same everywhere. Every unit of e-value in circulation must be matched, one for one, by real funds held in safeguarded accounts. The legal architecture, however, varies widely from one region to another. It determines who the contractual counterparty is, who is liable in a default, and what the funds may be used for.

• KYC is tiered, and that is what makes the model work. A tier 1 account opens with a phone number and a name, with very low balance and transaction limits. Higher tiers require an ID document, a photo, and sometimes proof of address, and they raise those limits. Designing a collection flow that ignores the customer’s tier means designing payment failures the merchant cannot explain.
• National ID becomes the keystone. Where a digital identity registry exists and can be queried, an account can be opened in a single session. Where none exists, tier 1 remains the de facto cap for a large share of users. The SIIPS 2025 report explicitly names this as one of the obstacles to scaling government and cross-border payments.
• Regulatory limits are product parameters. They vary by country, by tier, and sometimes by channel, and they change. A per-transaction limit, a daily limit, and a balance limit can each cause a legitimate payment to fail. Map them market by market and monitor them.
• The credit ban does not extend to partnerships. The overdrafts and nano-loans offered in the menus (short-term credit linked to the wallet) are booked by a partner bank or microfinance institution, not by the EME. The distinction is legal, not cosmetic: it determines who bears the risk and who is the lender under local law.

Interoperability: from bilateral deals to mandated switches

Interoperability is the ability of an account at one issuer to receive funds sent from an account at another. For a decade, African mobile money ran in silos. An M-PESA customer could send money to an Airtel Money customer only through SMS vouchers redeemed at an agent. Markets broke out of the silos along three separate paths: bilateral agreements negotiated by the market, a regulatory mandate, or a public switch. The path a country took largely determines the conditions under which a merchant can collect payments there.

Two patterns emerge from this history. First, deploying interoperability technically does not create usage by itself. Ghana’s GhQR standard (2020) is technically live but commercially disappointing, while TANQR took hold because TIPS membership made it mandatory in practice. The second pattern is about pricing. As long as an off-net transfer costs more than an on-net one, the largest player’s network effect stays intact, whatever the underlying infrastructure. The CBK and the BCEAO now intervene on fee schedules, not just on technical connections.

Collecting payments: merchant integration, costs, reconciliation

Collecting with mobile money means a merchant receives a payment into an e-money account held in its name. It requires opening a merchant account with each issuer, or with an aggregator that already has them. The merchant must also choose an initiation mode, wire up a callback, and design a reconciliation process that works even when responses go missing. Three initiation modes cover most of the market, and each has its own point of failure.

• Callbacks are unreliable by design. On congested mobile networks, a server notification may arrive late, twice, or never. The rule: treat every callback as idempotent (key = the issuer’s transaction ID), never treat a missing callback as a failure, and make periodic reconciliation by polling the source of truth.
• The issuer’s transaction ID is the only reconciliation key. Amount, timestamp, and phone number do not uniquely identify a payment. The same customer pays the same amount several times within a minute more often than you would expect.
• Settlement to the merchant is not instant. The till is credited immediately in e-money, but moving funds to a bank account, let alone repatriating them in foreign currency, follows a contractual cycle (often T+1 to T+2) and carries separate fees. This is the first item to negotiate, even before the collection fee.
• *There is no chargeback in the card sense. No multi-cycle dispute process, no network arbitration rules. Correcting an erroneous payment requires a reversal handled by the issuer on request, with highly variable turnaround times and success rates. A merchant therefore cannot rely on a standardized recourse mechanism: controls have to sit upstream*.
• FX and repatriation are a separate project. In several markets, the constraint is not collecting but getting the funds out: hard currency availability, central bank approvals, delays. Assess this constraint before signing, not after the first month of operations.

COLLECTION RECORD — fields to store every time

  psp_reference        ID generated by the integrator at initiation
  provider_txn_id      ISSUER's ID (M-PESA, MoMo, Airtel...)
                       -> RECONCILIATION KEY, unique, cannot be replayed
  provider             issuer + country (the same wallet differs by market)
  msisdn_masked        payer's number, masked at rest
  amount / currency    amount and LOCAL currency (never converted in the database)
  merchant_shortcode   till / paybill / merchant code used
  account_reference    order reference entered by the customer (often wrong)
  channel              ussd | stk | qr | app
  status               initiated | pending | success | failed | unknown
                       -> "unknown" IS A LEGITIMATE STATE, not a bug
  callback_received_at notification timestamp (may be null)
  queried_at           timestamp of the last status query
  settled_at           date of actual settlement to the merchant account
  fee_amount           fees charged, by amount tier

NON-NEGOTIABLE RULES
  1. idempotency on provider_txn_id (a callback can arrive twice)
  2. never set a final state based only on a missing response
  3. daily reconciliation against the issuer's statement, not
     against the internal ledger
  4. keep the provider_txn_id: it is the only evidence that holds up
     in a reversal request

Cross-border: remittances, corridors, and what PAPSS brings

Cross-border payments in Africa cover two distinct segments. The first is diaspora remittances sent to wallets, a mature, expensive segment driven by market competition. The second is intra-African interbank settlement, which long had to route through a correspondent bank outside the continent, in dollars or euros. That second segment is the target of a continental political project: PAPSS.

Sub-Saharan Africa remains the most expensive region in the world to send money to. The gap with the global average is about two percentage points on a $200 transfer. The cost depends mainly on the instrument the transfer originates from, and originating from a mobile wallet cuts it by more than half. That saving is why M-Pesa Global, MTN’s MoMo cross-border corridors, Zepz (WorldRemit and Sendwave), and Chipper Cash use the mobile wallet, not the bank account, as the landing point for funds.

• The heaviest corridor is not intra-African: it starts in the Gulf. The United Arab Emirates is one of the world’s largest remittance-sending countries (≈ $38.5 billion in 2023, No. 2 worldwide, World Bank), and about 88% of its population are expatriates. An operator targeting diaspora flows should study this corridor first.
• Currency unions change the nature of the problem. Within WAEMU, a “cross-border” payment among the eight countries is technically domestic: SICA-UEMOA (clearing) and STAR-UEMOA (RTGS) have operated under the BCEAO since 2004, and GIM-UEMOA is West Africa’s only multi-country domestic card scheme. The same logic applies in CEMAC with SYSTAC / SYGMA and GIMACPAY.
• Regional market share figures are often out of date. The latest public, cross-checked card market share figure for WAEMU dates from the end of December 2018: 28.87% for GIM-UEMOA versus 31.26% for Visa (BCEAO). Citing it without its date would be an analytical error.
• FX remains the opaque cost line. On wallet-to-wallet links and remittances alike, the rate applied and how transparent it is vary widely from one corridor to the next. The advertised fee is never the total cost: measure the all-in rate on a benchmark amount, corridor by corridor.

The Middle East, textbook failures, and blind spots

Mobile money in the Middle East looks different from mobile money in sub-Saharan Africa. It serves as fallback infrastructure in economies fragmented by conflict or by weak banking networks, rather than as a general substitute for bank accounts. Services there are more often anchored in the central bank than in the telecom operator.

Egypt combines both models in a single market. Mobile money there reached 55.5 million wallets and 1.4 billion transactions worth more than EGP 1,800 billion as of June 2025 (Central Bank of Egypt). Growth, however, comes mainly from the InstaPay / IPN instant bank rail, which has signed up more than 16 million users in three years. Financial inclusion in Egypt rose from 27.4% in 2016 to 76.3% in June 2025 (CBE). The telecom wallet was a first step toward electronic payments, not the end point.

• The next frontier is dormancy, not acquisition. With 347 million active accounts out of about 1.2 billion registered (GSMA, SOTIR 2026), the continent’s most accessible growth opportunity is not opening accounts but reactivating existing ones. SIIPS 2025 identifies government payments (G2P) and cross-border payments as the two levers to get there.
• Merchant payments are the real growth engine. They grew by nearly half to $155 billion worldwide in 2025 (GSMA, SOTIR 2026), and Safaricom reports 3.1 million merchants (+71%) in its FY26 fiscal year alone. This is the segment where payments professionals have the most to offer and face the least entrenched competition.
• New entrants keep coming. Four African services launched in 2025: Cashtel (Burundi), Bede (Sudan), Wave (Cameroon), and Gozem Money (Togo), per GSMA, SOTIR 2026. Watch Wave’s entry into CEMAC closely: the pricing model that upended WAEMU is now being tested in a market with mandatory interoperability.
• Never treat “Africa” as a single market in an investment committee. Sub-Saharan Africa alone has 173 live mobile money services, and the continent has 36 instant payment systems in 31 countries, four major currency zones, and as many e-money regimes. The right unit of analysis is the country, or at most the monetary union, never the region.