Nigeria’s informal economy moves faster than most digital tools can keep up with. Walk through Alaba International Market on any Tuesday morning, and you will see what that means in practice: hundreds of transactions completed before 9am, most of them logged nowhere except the seller’s memory and a scrap of paper tucked under the counter.
For years, the fintech industry’s answer to this problem was to build better apps. Cleaner interfaces, faster onboarding, cheaper smartphones. The assumption was that if you lowered the barrier to entry far enough, informal traders would eventually adopt structured record-keeping tools. That assumption has largely not held.
“The friction was never really about the interface,” says Onikola Waliyu, a Lagos-based fintech product architect whose work sits at an unusual intersection: the engineering of financial systems that are designed, from the ground up, to be invisible to the people using them. “Traders were already recording their activity. They were doing it in WhatsApp messages, in voice notes, in shorthand nobody outside the market could decode. The data existed. The problem was that nobody had built a system to catch it.”
A different kind of infrastructure problem
Waliyu spent roughly two years embedded in the operational reality of micro-SMEs before he began building what would become the SDCS Ghost Ledger. His conclusion, which runs counter to most fintech product thinking, was that the failure of digital tools in informal markets was not primarily a literacy problem or a cost problem. It was an architectural one.
Most accounting and ledgering software is built around a core assumption: that the person recording a transaction will stop what they are doing, open an application, navigate to the correct input field, and type structured data. In a high-velocity market environment, that assumption fails within the first hour of trading. A shop assistant managing three customers simultaneously is not going to pause to log a sale in QuickBooks.
“The tools were designed for people who have time,” Waliyu says. “Market traders do not have time. So the question I kept coming back to was: what if the system did not require any time at all?”
The answer he arrived at was a zero-interface ledgering architecture, a backend processing layer that sits inside WhatsApp, the platform Nigerian traders already use to communicate with customers, suppliers, and staff, and captures financial data from ordinary conversational messages without requiring any change in user behaviour.
The engineering challenge nobody had solved
Building such a system is technically harder than it sounds. The language informal traders use to describe transactions is neither standard English nor consistent Nigerian Pidgin. It is a fluid blend of regional shorthand, market slang, and abbreviated numerics that shifts by geography, by commodity sector, and sometimes by individual traders.
“Ten thousand naira might appear as ’10k’, or ’10K’, or ‘ten’, or just the number with no currency marker at all,” Waliyu explains. “A sale of two bags of rice might come through as ‘I don sell 2 bags for 15k’, or ‘rice x2 give Mama Ngozi 15’, or just ‘2 bags 15k’. Any competent human reader in that market understands all of those instantly. No off-the-shelf NLP system trained on financial data does.”
Victoria Fakiya – Senior Writer
Techpoint Digest
Stop struggling to find your tech career path
Discover in-demand tech skills and build a standout portfolio in this FREE 5-day email course
Waliyu’s solution was to engineer a custom parsing layer built on OpenAI’s GPT-4o-mini model, constrained by system prompts he designed specifically to handle the structural ambiguity of market Pidgin. The prompts force the model to return deterministic JSON output — extracting item names, quantities, amounts, and customer identifiers — regardless of how the input is phrased. Across internal testing benchmarks comparing raw message inputs to parsed outputs, the system has maintained above 95% accuracy at a response latency of under 1.5 seconds.
That latency figure matters as much as the accuracy one. WhatsApp interactions feel instant to users. A confirmation message that arrives three or four seconds after a sale has been sent breaks the conversational rhythm and trains users to ignore the system. Waliyu engineered the entire request-response loop, from incoming Meta Cloud API webhook to parsed database entry to outbound confirmation, to complete within that 1.5-second window, a constraint that shaped several significant architectural decisions downstream.
Deliberate architecture, not default choices
One of the more technically instructive decisions in the Ghost Ledger’s design is one that most developers would not have made. For the owner-facing analytics dashboard, Waliyu chose Jinja2 Server-Side Rendering rather than a React or Vue-based Single Page Application.
In most professional frontend contexts, that choice would raise questions. React is the industry standard for this class of dashboard application, and for good reason. But Waliyu had data from early pilot deployments showing that React-based dashboards were consistently failing to initialise on low-end Android devices operating on dropping 3G connections in dense market environments. The issue was not network speed per se, but the volume of client-side JavaScript that React requires the browser to parse and execute before anything appears on screen.
Server-Side Rendering offloads that work to the server before transmission, delivering a pre-rendered HTML page that displays immediately regardless of the device’s processing capability. For a trader checking their daily revenue on a three-year-old Tecno handset with intermittent connectivity, the difference between a dashboard that loads in half a second and one that times out is the difference between using the product and abandoning it.
“I tested both approaches with actual users in actual market conditions,” Waliyu says. “The React version had better code. The SSR version had better outcomes. I chose the SSR version.”
Security architecture received the same level of considered attention. Every incoming message processed by the Ghost Ledger passes through an HMAC SHA-256 verification handshake before it reaches the parsing layer, a protocol that authenticates the message source against Meta’s Cloud API signature and rejects any payload that does not match. Waliyu implemented this before it became a standard requirement, reasoning that a system whose output would eventually be used as income evidence in formal bank applications needed to establish data integrity at the point of ingestion, not as an afterthought.
What the numbers show
The Ghost Ledger entered limited deployment in early 2025. As of July, the system is actively used by more than 1,500 micro-SMEs across Lagos and its surrounding markets.
“These are not pilot users. They are not people we onboarded through institutional partnerships or incentive schemes. They found the system, tried it, and kept using it because it fit into how they were already operating.”
Retention in informal-market fintech products is notoriously difficult to achieve. Tools that require behavioural change see a steep drop-off within the first month. The Ghost Ledger’s adoption curve has been different, and Waliyu attributes that to the zero-interface design principle: because the system does not ask traders to do anything they were not already doing, there is no new habit to form and therefore nothing to abandon.
The larger implication
The Ghost Ledger is not, at its core, a product story. It is an infrastructure argument.
The financial invisibility of informal traders is not simply a record-keeping problem. It is the reason that businesses turning over millions of naira monthly cannot access formal credit, cannot demonstrate operational stability to potential partners, and cannot scale beyond the limits of what their owners can personally track and remember. The system Waliyu has built does not ask traders to become more formal. It makes their existing activity legible to formal systems.
Independent fintech analysts have noted that this approach, meeting informal markets at the point of their existing behaviour rather than asking them to adopt new ones, represents a meaningful shift in how financial inclusion products are being designed across the continent.
“The credit history problem has always been framed as a data problem,” says one David Inem, a Lagos-based analyst who reviewed the system. “What Waliyu recognised is that it is actually a capture problem. The data is there. It just needs a system intelligent enough to catch it.”
Whether the Ghost Ledger becomes the infrastructure layer that finally bridges Nigeria’s informal economy with its formal financial system remains to be seen. What is already visible is that Waliyu built something technically serious to solve a problem that the industry had largely written off as unsolvable, and that a meaningful number of traders are now using it to run their businesses.
That, in itself, is worth paying attention to.
