M-Pesa's export problem: why the world's best mobile money didn't travel

M-Pesa is the most successful mobile-money system ever built — and its export record is one of the most instructive failures in emerging-market tech. In Kenya it became money itself: by 2012 it had 19.5 million users (about 83% of the adult population), and by 2015 roughly 43% of Kenya's GDP flowed across it. Then Safaricom and Vodafone tried to take it abroad — and in the markets that mattered most for the "just export it" thesis, it didn't take. Vodacom shut M-Pesa in South Africa on 30 June 2016; Vodafone closed it in Albania in July 2017. The lesson isn't "the tech didn't work." The tech was never the moat. M-Pesa won on things that are stubbornly local — a dense human agent network, a trust anchor in Safaricom's telecom dominance, a one-time regulatory window, and a large unmet pain. None of those travel. Export the app without the local substrate that made it win, and you get a great product nobody adopts.

What they did — and the going-global move

M-Pesa launched in March 2007 as a Safaricom product for sending money by SMS. Its real infrastructure was human: the agent network grew from about 450 outlets in mid-2007 to ~18,000 by 2010 and ~40,000 by 2016 — corner shops that turn cash into digital value and back. That density, not the app, was the moat.

The going-global thesis was seductive: we cracked mobile money, export it. Vodafone and Safaricom pushed M-Pesa across Africa and into Europe and Asia — including South Africa, Romania (its first European market, 2014), Albania, and India. The outcomes split hard.

The teardown — through the deeper-local lens

Treat M-Pesa as a product and you miss it; it was a local system. Four local primitives made it win, and each failed to transfer:

1. Agent density. You can't buy 40,000 trusted cash-in/cash-out points overnight. You rebuild them, market by market — or you don't have M-Pesa, you have an app.
2. Trust anchor. Kenyans trusted M-Pesa because they trusted Safaricom. In South Africa the trust anchor was banks, not a telco — the same product arrived with no borrowed credibility.
3. Regulatory window. Kenya's regulator let a telco move money before the banking lobby could close the gap. Elsewhere that gap was already shut.
4. The pain shape. Kenya had a huge unbanked population and a strong internal-remittance need. South Africa didn't: financial inclusion was already high, so the pain M-Pesa solved was smaller — and it showed. Vodacom's own numbers were brutal: just 76,000 sign-ups in 2015, nowhere near the critical mass the model needs. It pulled the plug in June 2016; Albania followed in 2017.

Same product; opposite results — because the substrate was different.

My operator take

The un-obvious lever isn't the wallet UX; it's the trust-and-distribution substrate, and it is almost always local. Before you enter a market, split your advantages into two buckets: portable product (the app, the ledger, the brand) and local substrate (agent density, the specific trust anchor, regulatory timing, the exact pain shape). M-Pesa's moat was mostly substrate — and you cannot ship substrate. You rebuild it, or you decline the market. The teams that win glocal don't "expand"; they re-found, locally, each time.

The playbook takeaway

A pre-entry checklist for any locally-dominant product going abroad:

• List your top five advantages. Tag each portable or local-substrate.
• For every substrate advantage, write the plan to rebuild it — agents, trust anchor, regulatory path, pain fit. No plan, no entry.
• Kill the assumption that your home moat travels. It usually doesn't — that's the point.
• Pick markets by substrate-rebuild cost, not market size. High financial inclusion (South Africa) is a lower opportunity for a substitution product, not a higher one.

This is a spoke in the Going Glocal cluster — the un-obvious local things (trust, distribution, regulation, money movement) that the frameworks skip.

One honest open question

Is there a clean case where local substrate did travel — a locally-won advantage that transferred intact — or is "re-found locally" simply the law of glocal expansion? Kenya's own regional M-Pesa spread (Tanzania, the DRC) looks less like a transfer and more like a rebuild where the pain shape rhymes. I don't yet have a counterexample where it moved without one.