There is a persistent misconception about what AI does in trade finance decisions. People assume it works like a credit score — pull a number, apply a threshold, approve or decline. That model works reasonably well for personal loans or consumer credit, where the primary variable is the individual borrower's repayment history. Trade finance is more complex. You are not lending to a person; you are lending against a transaction. And transactions have their own risk profile that is entirely separate from — and sometimes more predictive than — the borrower's balance sheet.
When we built the scoring model at Trade Lenda, the first thing we had to decide was: what are we actually scoring? The answer is not the importer. The answer is the trade: the specific shipment, from a specific origin, of a specific commodity, to a specific destination, involving a specific overseas counterparty, through a specific route and port corridor. Each of those variables carries independent risk. The job of the model is to read all of them simultaneously and produce a composite risk signal before a human underwriter opens the file.
Shipment Risk Is Multidimensional
When traders and bankers talk about "shipment risk," they often mean one thing: will the goods arrive? That is too narrow. In West African trade corridors, shipment risk has at least five dimensions that each need to be assessed independently.
Origin port and country risk
The origin country's export controls, documentation reliability, and transit time reliability all affect the risk profile of a shipment. A container shipped from Guangzhou via a direct vessel call to Apapa carries a different risk profile than the same commodity shipped from Lagos to a re-export hub in Lomé and then re-forwarded. The routing matters because it affects documentation chain integrity, transit time predictability, and the likelihood of goods matching their declared manifest by the time they arrive at NCS inspection.
We track origin-port pairing risk across the major sourcing corridors for Nigerian imports: China (Guangzhou, Ningbo, Shanghai), India (Nhava Sheva), UAE (Jebel Ali), and Turkey (Mersin). Each corridor has a distinct risk profile for document reliability, average transit time variance, and commodity misclassification rates — patterns that emerge from the NCS clearance data over time.
Commodity category and HS code risk
Some commodity categories are inherently higher-risk in Nigerian customs, not because of the goods themselves, but because of how they are commonly misdeclared. Electronics components, used machinery, and certain chemical precursors have historically higher rates of HS code misclassification — either because importers genuinely misunderstand the tariff schedule or because they are attempting to reduce duty liability. Our model weights commodity HS codes against their historical clearance flag rates at NCS.
This is distinct from saying those goods are bad to finance. We are not saying an importer bringing in electronics from Guangzhou is a problem. We are saying the commodity category has a higher variance on clearance time, which is a risk to the capital timeline, and that affects the risk score accordingly. The importer can have a perfectly clean customs history and still see a slightly elevated commodity risk score if the HS code has a historically volatile clearance profile.
Route and transit time risk
Transit time variance directly affects capital risk. If we advance working capital with a 45-day tenor against a shipment that takes 25–30 days from Guangzhou to Apapa under normal conditions, a 15-day Suez Canal or port congestion event leaves a very tight repayment window once goods clear and sales are made. Our model incorporates live and historical transit time data for the major routes into Lagos and Tin Can Island port, including seasonal congestion patterns. The July–September West Africa monsoon season, for example, historically compresses vessel scheduling flexibility at certain ports.
Port congestion and dwell time risk
Apapa is the largest port in sub-Saharan Africa by volume, and it runs near capacity for significant parts of the year. Dwell times — the period between vessel arrival and cargo clearance — are not just a function of the importer's documentation speed. They are also a function of the terminal's operational state, truck access conditions on the Apapa access road, and NCS scanning queue backlogs. We pull vessel arrival and departure data, terminal dwell-time averages, and NCS inspection queue estimates to assess the realistic clearance timeline for a specific shipment. A trade that looks clean on paper but is arriving during a peak congestion period gets a timing-risk flag that adjusts the advance tenor offer accordingly.
Commodity price volatility
For commodity-sensitive imports — raw materials, certain agricultural inputs, bulk chemicals — the value of the goods at the time of repayment may differ from the value at the time of advance. A sharp drop in commodity prices between advance and repayment compresses the importer's margin and increases default risk. Our model tracks commodity price index movements for the major categories we finance and flags trades where there is material price volatility risk within the expected tenor window.
Why Buyer Payment History Matters for Import Transactions
This one is less intuitive. For an import transaction, the importer's overseas supplier is paying nothing to Trade Lenda — we advance capital to the Nigerian importer. So why does the overseas buyer's payment behavior matter?
It matters because for many Lagos importers, they are not buying for their own consumption. They are buying to resell — to distributors, retailers, or end customers who are themselves a counterparty. For export transactions, the overseas buyer's payment reliability is directly relevant because the repayment of our advance depends on the buyer settling the export invoice. But even for import transactions, the importer's downstream buyer history — how reliably their commercial customers pay — affects their working capital cycle and their ability to repay trade finance on schedule.
We also use cross-corridor buyer history to build a picture of how specific overseas suppliers treat Nigerian buyers: whether payment terms are honored, whether pre-payment demands are common, and whether there have been supply shortfalls or document disputes. This is intelligence that helps us assess whether a specific trade is likely to run smoothly or encounter friction that delays repayment.
The Composite Score: What It Means and What It Does Not
The output of our risk model is a composite score and a set of flags. The score is a single number — we show it as a value out of 100 in our internal review interface. Higher scores indicate lower compound risk across all the dimensions described above. The flags are specific items that our human underwriter needs to review before making a final decision: a commodity category with elevated clearance variance, a transit routing with recent congestion events, or a buyer payment record that has gaps.
We want to be precise about what the score does not mean. It does not mean we will automatically approve a trade above a certain threshold. It means we have a pre-structured view of the risk that allows the underwriter to spend their time on the specific flags rather than re-doing the data assembly from scratch. A trade with a score of 88 might still be declined if one of the flags represents a risk the underwriter judges as unacceptable given the tenor and advance amount. A trade with a score of 71 might still be approved if the underwriter can verify that the flagged item is a data artifact rather than a genuine risk.
We are also careful about what the model cannot see. Our scoring does not include information about the importer's broader business finances beyond their customs and trade history. We do not currently model political risk at sub-national level within Nigeria — the model treats Nigerian destinations uniformly. And we do not have visibility into informal side agreements between importers and their customers that might affect repayment behavior. These are genuine limitations, and our underwriters know them.
How the Model Learns From West African Trade Corridors Specifically
One of the reasons we built this model from the ground up rather than adapting a generic trade finance scoring tool is that West African trade corridors have risk characteristics that generic models do not capture. The Apapa–Cotonou re-export dynamics, the seasonal pattern of agricultural commodity imports from the Sahel into Lagos, the specific clearance friction patterns at Tin Can Island versus Apapa for certain commodity categories — these are details that only appear in data built on actual Nigerian customs records and West African shipping patterns.
Our model is trained and calibrated on trade data from the corridors we operate in. We update it as new patterns emerge. The 2023–2024 Naira devaluation cycle, for example, created new commodity price volatility patterns that our model was initially underweighted on and that we recalibrated to reflect. Building and running the model is not a one-time exercise. It is ongoing work that requires people who understand both the technology and the specific trade environment it is scoring.
What This Means for Importers Applying for Finance
If you are a Nigerian importer applying to Trade Lenda, the AI scoring step is not an obstacle. It is what makes the 48-hour decision possible. The model does the data assembly — your NCS history, your commodity risk profile, the shipment route, the timing against port conditions — and surfaces the result to a human underwriter who reviews it with judgment. That combination of automated data processing and human review is faster than either one alone.
The practical implication: importers with consistent, well-documented customs histories and trades in lower-risk commodity categories will see faster, cleaner decisions. Importers whose trades involve higher-variance commodities or corridors will see more questions from the underwriter — not necessarily a decline, but a deeper review. We will always tell you what we are looking at and why. See the full process here.