Between 2020 and 2023, the most funded B2B marketplace sector in African tech history quietly fell apart.
Hundreds of millions of dollars had flowed into platforms promising to digitize informal retail across Nigeria, Kenya, and East Africa. Investors were convinced that the combination of commerce and embedded credit was the unlock. That is, platforms sitting on merchant transaction data could lend against it more accurately than any bank ever could. The thesis was sound. The execution, in too many cases, was not.
At least two major players suffered significant losses on embedded lending and paused it entirely. For example, RejaReja, MarketForce's B2B marketplace, faced loan repayment difficulties from shopkeepers that directly damaged the company's cash flow.
MarketForce's credit partner, Pezesha, eventually filed a liquidation petition over unpaid debts, stemming from a cash crunch that made it impossible for the company to meet its monthly obligations. The platform had served 270,000 merchants across five countries. It shut down entirely in 2024.
What makes this story instructive is not the failure itself, African tech has always had failures, but the contrast sitting right beside it. While RejaReja was shutting down, OmniRetail, a Nigerian B2B commerce platform, was building the same combination of FMCG distribution and embedded credit, becoming EBITDA-positive in 2023 and net profitable in 2024.
Its embedded finance arm, OmniPay, processed over ₦1.3 trillion ($810 million) in transactions in 2024 and now disburses ₦19 billion monthly in BNPL inventory credit, achieving near-zero defaults.
Same market, same product category, and same thesis, but completely different outcomes. Understanding what separated them is not just interesting; it is the playbook every African platform considering embedded lending needs to read before building.
The Scale of What Is at Stake
Before getting into the failure analysis, it is worth establishing what the opportunity cost of getting this wrong looks like.
Africa's embedded finance market is projected to grow from $11.9 billion in 2024 to $18 billion by 2030, with embedded credit products becoming a default financing route for small enterprises across Kenya, Nigeria, Ghana, and South Africa. Plus, Nigeria's embedded finance market alone is expected to grow at a CAGR of 25.5% from 2024 to 2029, reaching $3.48 billion by 2029.
The advantage of building embedded lending well for platforms in the next three years will be accumulated merchant financial data, repayment track records, and credit relationships that will be extremely difficult for new entrants to replicate.
On the other hand, not building it well will set their products back years and, in the worst cases, destabilize the underlying commerce business that the lending was supposed to strengthen.
The stakes run in both directions. Which makes understanding the failure patterns not academic but operationally urgent.
Why Embedded Lending Fails: Five Patterns That Keep Repeating
1. Underwriting on the Wrong Data
One of the most enticing things about embedded lending from a platform's perspective is the transaction data. A marketplace that sees a merchant ordering stock every week, paying on time, and growing order volumes has something a bank does not: a live, continuous record of commercial behavior. That data feels like a complete credit picture. It is not.
Platform transaction data reveals behavior within your ecosystem. It tells you nothing about the merchant's obligations outside it: other lenders they are repaying, rent they are carrying, and informal loans they took last month to cover a cash-flow gap. A merchant who looks financially healthy on your platform may already be stretched to the point where any new obligation tips them into default.
The pivot to transaction-data-based underwriting across African platforms has exposed the gap between platform-native data and the full financial picture. MSME financing gaps and the absence of collateral have forced platforms into alternative credit scoring, but the quality of that scoring depends entirely on how complete the data inputs are.
The platforms that have solved this are the ones that combine their own transaction data with Open Banking-sourced bank account data, giving them visibility into the merchant's full financial life, not just the slice that runs through their platform. Underwriting a merchant on partial data is not a conservative approach. It is a way of approving a risk profile you cannot see.
2. Repayment Infrastructure Built as an Afterthought
Building the loan product is the part that gets attention. Building the repayment infrastructure is the part that determines whether the product survives.
RejaReja's loan repayment difficulties from shopkeepers directly impacted MarketForce's cash flow, which means the platform was running credit at scale without the infrastructure to reliably collect it. Manual repayment processes, fragmented payment channels, and collections that depend on merchant goodwill rather than automated systems are not sustainable at any significant loan book size.
The failure pattern here is not malice or even poor product design, but a sequencing problem. Platforms launch the credit product because it improves conversion and order values. They figure out collections when defaults begin to arrive. By then, the loan book has grown faster than the collections infrastructure can keep up with.
The correct sequence is the reverse: build the repayment architecture before the loan product goes live. Consent-based direct debit mandates, in which merchants agree to automated repayment from their accounts on scheduled dates, remove human follow-up entirely from the collections equation. Repayment happens whether or not the merchant remembers to initiate it, and the platform's collections cost drops to near zero.
3. Treating Embedded Lending as a Product Feature Rather Than a Business Line
Many platform teams approach embedded lending the way they would add a new product category to their catalog, as an incremental feature requiring additional resources. The reality is that lending is a fundamentally different business from commerce or logistics, and the governance, risk management, and compliance infrastructure it requires is not incremental.
Standalone lenders that lack platform reach may retreat or become infrastructure providers. The margin and risk premium on embedded credit depend on the quality of integration, underwriting precision, and partner alignment. For platforms that embed credit, the inverse is equally true: embedded credit without lending-quality infrastructure becomes a liability rather than an asset.
The platforms that treat embedded lending as a feature ship fast and learn slowly, through defaults, regulatory attention, and reputational damage with the merchants they were trying to serve. On the other hand, the ones that treat it as a business line build dedicated risk functions, invest in underwriting data quality, and approach compliance as a design requirement rather than a post-launch consideration.
4. Ignoring the Compliance Architecture Until It Is Too Late
Nigeria's regulatory environment for embedded lending is now among the most specific in Africa, and most platform teams are not fully across what it requires of them.
The FCCPC's DEON Regulations, which took effect in July 2025, extend to every participant in a lending value chain, not just licensed lenders, but the platforms that distribute or facilitate consumer credit.
Therefore, if your agritech app facilitates input financing, your marketplace enables merchant credit, or your logistics platform offers driver loans, you are covered, and the obligations are specific. Mandatory FCCPC registration, partnership agreement submission for approval before going live, annual compliance reporting, and prohibitions on accessing borrowers' contact data or using aggressive collection methods.
The businesses that discover these obligations after they have already launched a credit product face a harder problem than the ones that build compliance in from the start. Retrofitting a consent architecture, a data governance framework, and a collection conduct policy onto a live product is significantly more expensive and more disruptive to the user experience than building them in at the design stage.
5. Misaligning Loan Terms With Borrower Cash Flow Cycles
This failure pattern is specific to African markets and consistently underestimated by platform teams who have not studied their borrowers carefully enough.
A market trader whose revenue peaks on weekends, a farmer whose income arrives in one or two harvest settlements per year, a logistics driver whose earnings fluctuate daily, none of these borrowers has the fixed monthly cash flow that standard loan repayment schedules assume. When repayment dates land at the wrong point in a borrower's earning cycle, default is not a sign of bad intent. It is a consequence of bad product design.
Therefore, the platforms that get repayment right structure loan terms around the actual cash flow patterns of their specific borrower population. Farmers get repayment at harvest. Merchants get collections aligned to their sales settlement cycle. Drivers get micro-deductions from each completed job rather than monthly lump sums.
This requires knowing the borrower's financial patterns well enough to design for them, which brings us back to the data question at the beginning.
The Infrastructure That Separates the Two Outcomes
Every failure pattern described has a corresponding infrastructure solution. The platforms that fail at embedded lending are almost always the ones that try to build that infrastructure themselves, underestimate how long it takes, and ship before it is ready. On the other hand, the ones that succeed are increasingly those that integrate existing infrastructure rather than build from scratch.
This is why we built Zeeh, starting with the Connect, which addresses the underwriting data problem directly by pulling verified transaction data with the borrower's explicit consent through Nigeria's Open Banking framework, giving platforms the full financial picture of their borrowers rather than just the slice that runs through their own systems.
The income verification, expense pattern analysis, and visibility into existing obligations that proper credit underwriting requires are all sourced from the bank account in real time, not from submitted documents that can be manipulated.
Then, Insights converts that raw transaction data into structured credit intelligence, income categorization, cash flow stability scores, affordability calculations, and risk signals that flag financial stress before it becomes default.
For a platform building credit products for informal merchants or farmers, this is the difference between underwriting decisions grounded in evidence and approvals based on incomplete proxy data.
ZeehID handles the identity verification layer, including biometric matching, document verification, liveness detection, and real-time BVN and NIN database checks, all of which meet the CBN's May 2026 mandatory standards. Embedded lending begins with identity, and identity verification that fails to meet the regulator's current requirements is not a technical shortcoming; it is a compliance exposure.
Finally, Direct Debit closes the repayment loop through consent-based, automated, mandate-based collection aligned with the borrower's actual payment cycles, and is built to meet the FCCPC's 2025 requirements. For platforms whose embedded lending experiments have previously collapsed at the collections stage, this is the infrastructure that makes repayment reliable rather than dependent on borrower initiative.
For a broader picture of where embedded finance is heading across African markets — the five verticals, the regional dynamics, and how the infrastructure layer is evolving — our analysis of embedded finance in Africa covers the full landscape. And if you are building embedded lending specifically into a marketplace, the embedded finance playbook for African marketplaces walks through the complete infrastructure and compliance requirements.
The Bottom Line
The platforms that have failed at embedded lending in Africa were not building bad products. They were building good products on incomplete foundations, partial data, manual repayment, compliance gaps, and loan terms misaligned with how their borrowers actually earn.
The platforms that have succeeded have built the foundation first. Successful businesses did not achieve near-zero defaults because their merchants are unusually disciplined.
They achieved them because the data was complete, the repayment was automated, the terms fit the cash-flow cycle, and the product was integrated deeply enough into the merchant relationship that repayment was not a separate obligation; it was just part of how the business moved through the platform.
Africa's embedded finance competitive landscape is shifting toward infrastructure consolidation, with API providers and platform enablers playing a pivotal role. The competitive edge will lie in contextual product design, data-enabled underwriting, and cross-sector integration.
The platforms that own that edge in their specific verticals will carry lending advantages that compound over time. The ones still trying to build the foundation from scratch in 2026 are already behind.
Embedded lending in Africa works, especially when it’s built properly. Talk to the Zeeh team about building embedded lending infrastructure that works from day one.
