Stop buying users you will never keep.

Look at the customer-acquisition mathematics of consumer financial apps across East and West Africa over the past five years, and you can watch Silicon Valley growth metrics destroy hundreds of millions of dollars of shareholder value in real time. Egged on by growth-stage investors demanding exponential monthly-active-user charts, consumer fintechs waged a war for mass-market retail wallets: free $100 point-of-sale terminals for roadside merchants, subsidised zero-fee transfers, $1 to $2 cash referral bonuses for every new download. The playbook was imported intact from markets where capital was free, and fraud was an edge case. In its new habitat, both assumptions failed.

Boardroom spreadsheets missed a basic truth of emerging-market economics: when a $1.50 referral bonus equals half a day's casual wage in an economy where four out of five workers are informal, you do not create loyal banking customers. You create a professional arbitrage industry. Across major African cities, organised street-level syndicates learned to batch-register thousands of SIM cards using borrowed or forged identity documents, harvest the referral payouts, cash out at the nearest till and abandon the app forever. Venture investors got a hockey-stick download slide for the next fundraising round while the income statement absorbed a permanent cash furnace. When the global rate cycle turned, and the funding winter arrived, the furnace kept burning; only the fuel stopped.

Separating the referral tourist from the unit-cost ledger.

The same unit-economics illusion plagues merchant acquiring. Suppose a payments institution gives a retail merchant a KSh13,000 smart terminal for free and undercuts competitors by charging a 0.4 per cent merchant discount rate. Deduct switch fees, cellular data on the terminal, field-agent maintenance, and cost of capital, and the acquirer retains less than 10 basis points of net margin. The merchant must process KSh13 million in sales just for the provider to recover the hardware, yet the moment a rival sales agent walks in offering a slightly cheaper rate or a promotional giveaway, the first terminal goes into a drawer and the new one gets plugged in. Hardware subsidies rented the counter; they never bought the merchant.

No institution can profitably bank a low-income mass market by subsidising expensive hardware to acquire transient users carrying KSh300 average balances. The operators who survived the 2023 to 2026 funding winter did three things: walked away from the subsidised mass-market street fight, refocused direct retail propositions on mid-to-upper-tier customers with genuine transaction velocity, and opened their underlying regulatory and settlement plumbing as business-to-business infrastructure for other companies to rent.

Renting the regulated pipe instead of pouring concrete for three years.

Think about what a regional airline, a consumer-goods distributor, a horticultural exporter or a Tier-3 bank needs when it wants to embed wallets, virtual collection accounts or cross-border supplier payouts into its workflow. Building that stack in-house requires three years of licensing across multiple central banks, core ledger engineering, switch certifications and a dedicated compliance team. Renting a white-label B2B API from an established infrastructure provider lets the enterprise launch in six weeks, converting a heavy capital-expenditure gamble into a predictable variable cost per transaction. For the infrastructure provider, powering three hundred corporate balance sheets, rival consumer fintechs included, yields sticky, contractually locked volume with zero retail marketing burn.

Evaluating an embedded-finance partnership still requires checking how each technical layer fails. Deterministic API routing rules can split an incoming distributor payment automatically between tax withholding, logistics fees, and net supplier credit at near-zero cost. A fraud model on the B2B gateway will misfire if trained on retail consumer behaviour rather than lumpy payroll and inventory cycles. Developer copilots can compress a client's integration from months to days, so long as human engineers audit the production authentication code. And any automated treasury script sweeping corporate float across partner banks must sit behind strict reconciliation gates, so a webhook timeout cannot duplicate a multi-million-shilling batch payout. Often the cheapest first step for a corporate treasurer is simply consolidating fragmented collection paybills into a single reconciled host-to-host bank API before buying any new software at all.

Auditing customer acquisition against twelve-month gross margin.

Before signing off the 2027 growth budget, the chief executive and chief financial officer should run a cohort audit across every retail and merchant acquisition channel, stripping out gross registered-user counts to measure 90-day un-incentivised retention, alongside the months required to recover fully loaded acquisition and hardware costs from net gross profit. If any retail segment needs more than twelve months of realistic net transaction margin to repay its onboarding subsidy and referral cost, kill the subsidy immediately and reallocate the capital into B2B workflow APIs where enterprise clients pay for utility on day one.

Then ask the question that separates a franchise from a fundraising deck. In your 2027 budget, how many of your active users would still transact next Tuesday if you switched off every promotional subsidy and fee waiver tonight?