The entire subscription economy is built on one quiet assumption: once a customer hands over a card, you can charge it again next month without asking. Card-on-file is the engine. Now take that assumption to markets where the dominant way to pay isn't a card at all, and you find the real reason subscription businesses struggle with mobile money.
push by default
Mobile money is push by default. The customer initiates every single payment, usually by approving a prompt on their phone. That's a wonderful property for one-off purchases and a terrible one for subscriptions, because a renewal that depends on a human tapping "approve" every month isn't a renewal. It's a monthly re-sale to someone who has already bought.
So the question that matters, network by network, is this: does the network let a customer grant standing permission, and once granted, who pulls the trigger each cycle? In practice you find three models.
model one: the customer-side standing order
The first is the customer-side standing order, what a South African would call a stop order. The customer instructs the network to pay a merchant a set amount on a schedule. M-Pesa has a product in exactly this shape. It's genuine recurring, but notice who owns it: the customer sets it up, and the amount is fixed at creation. Fixed-amount mandates are fine for fixed-price plans and painful for anything variable, usage-based or priced against a moving exchange rate.
model two: merchant-initiated pre-approval
The second is the merchant-initiated pre-approval. The customer approves a mandate once, and the merchant charges against it each cycle, within the bounds the mandate allows. MTN MoMo has a construct in this family. This is the closest mobile money gets to card-on-file, and where it's available and properly supported, subscriptions genuinely work.
model three: not actually recurring
The third isn't really recurring at all: a fresh payment prompt every cycle, dressed up in recurring clothes by whoever is selling it to you. Every renewal is another interruption on the customer's phone, and every interruption is a chance to churn. If a provider claims recurring support, this is the first thing to check, because a lot of what's marketed as recurring is model three wearing a suit.
verify before you build
Coverage across networks is uneven, the details differ by market, and anything you plan to build a business on deserves written confirmation from the network itself rather than a diagram in an aggregator's pitch deck. That's not cynicism, it's just how infrastructure diligence works. We hold ourselves to the same standard.
what this means for pricing and dunning
Two design consequences follow if you're a subscription business looking at these markets. Price in local currency at stable price points wherever the mandate model fixes amounts, because FX-linked pricing and fixed mandates don't mix. And rethink dunning entirely: there's no card to retry, so recovery is about wallet balance and timing, which makes payday-aware retries and respectful reminders over the channels people actually read worth more than any retry algorithm from the card world.
None of this is a reason to avoid mobile money. It's most of the continent's spending power, and recurring on it is absolutely achievable. It just isn't one integration and one behaviour. It's per-network engineering, and that's precisely why so few have done it properly.