A consumer applies for a short-term loan. Your bureau pull shows a 670 score, three existing obligations, total monthly commitments of R4,200 on a R15,000 salary. The discretionary income looks adequate. You advance the credit. Six weeks later, the consumer misses the first instalment. When you pull the bureau again, you now see two additional loans that were not there before, both taken out within the same fortnight as your application. The stacking happened before you lent. Your bureau just didn't know yet.
This is not an edge case. The NCR Consumer Credit Market Report Q1 2025 recorded 18.08 million credit applications in the quarter. With SA lenders operating across dozens of NCR-registered platforms, the mechanics of the bureau reporting cycle mean there is a structural window between when a loan is disbursed and when it appears on bureau data. That window is where loan stacking happens.
What loan stacking is and why it happens
Loan stacking describes a consumer taking multiple credit agreements from different lenders within a compressed time period, typically within the same month or even the same fortnight. The motivation varies. In most cases it reflects genuine financial distress: a consumer who has exhausted one credit option uses the disbursement from a first loan to meet an urgent obligation, then immediately takes a second loan to replace the funds, and so on. In a smaller proportion of cases, particularly in micro-lending, stacking is deliberate fraud. The consumer never intends to repay and is extracting maximum disbursements before the system catches up.
The bank statement signature of stacking is distinctive. Multiple income-type credits from different financial services providers appearing within a two-to-three week window. Multiple new debit orders to unfamiliar payees (names that look like lending platforms rather than insurers or utilities) appearing in the same period. Cash withdrawal patterns that are consistent with receiving disbursements and immediately converting them into non-traceable outflows. Behavioural credit scoring captures all of these signals from transaction categorisation. The bureau captures none of them in the window between disbursement and submission.
The bureau data lag: why you won't see it until it's too late
SA credit bureaus operate on monthly submission cycles. Lenders submit their updated portfolio data to the bureau at intervals that may be four to six weeks after the payment or origination event. A loan disbursed on 3 June will not appear on the consumer's bureau record until the lender's July submission cycle closes, which may be 5 July or later.
For a consumer stacking loans across five different micro-lenders in June, every lender who pulls the bureau during June is looking at pre-stacking data. The consumer appears to have three obligations. They actually have eight by the time the last stacking event closes. By the time the bureau reflects the full obligation picture, every lender has already advanced their portion.
This is not a failure of the bureau system as designed. Monthly submission cycles are the established operating model across SA bureaus. It is a structural limitation that creates a predictable risk window. Bank statement analysis is specifically positioned to fill this gap because the bank statement reflects disbursements and new debit orders in real time, before the bureau cycle closes.
What Regulation 23A requires that the bureau can't provide
Regulation 23A requires that a credit provider assess the consumer's “all monthly debt repayment obligations,” not merely those already listed on the bureau. The distinction is legally material. A credit provider who assesses obligations based only on the bureau pull, when a bank statement would have revealed additional recent loans, may not have taken the “reasonably practicable steps” that Section 81(3) requires for income and obligation verification.
The NCA's affordability assessment framework is designed to prevent loan stacking. The prevention mechanism is the Section 81 affordability assessment obligation. But that obligation is only effective as a stacking prevention tool if the assessment captures current obligations, including those not yet on bureau. A bureau-only assessment cannot satisfy the “all monthly obligations” standard for a consumer who has stacked loans in the current cycle.
The consequence for credit providers is that relying solely on bureau data when a bank statement was available, and the bank statement would have shown stacking signals, may not constitute a complete defence against a reckless lending challenge. The obligation to verify extends to information that was reasonably obtainable at origination.
The bank statement signature of loan stacking
Transaction analysis distinguishes stacking from a consumer with multiple long-standing obligations. The stacking signature is temporal compression: multiple new financial transactions appearing within a short window, from providers who were not present in earlier months of the statement.
The primary indicators in order of reliability:
- Multiple income-type credits from financial services names: Salary appears on the 25th. Then on the 3rd, a deposit from “Boodle Loans”. Then on the 8th, from “Lime24”. Then on the 12th, from “FairMoney ZA.” Each appears as an income credit. None were present in months one and two of the statement.
- New debit orders to unfamiliar payees: A debit order from “Capital Credit” starts in month three. A second from “MFC Services” also starts in month three. Neither has a history in prior months. The rate of new debit order origination, particularly from financial-services-category payees, is itself a stacking indicator.
- Cash withdrawal pattern change: A consumer who previously withdrew R1,200–R1,800 per month in ATM cash begins withdrawing R4,500–R6,000 per month. High cash withdrawal ratios correlate with informal debt servicing, such as payments to mashonisa lenders or informal arrangements that never appear as structured transactions.
None of these signals are definitive in isolation. A consumer who receives a bonus, starts a side business, or consolidates existing debt will show some of these patterns. The risk assessment is pattern combination and temporal compression, not single-signal triggering. Three new financial-services debit orders appearing in the same month alongside multiple income-type credits from different providers suggests a meaningfully higher-risk profile than one new debit order appearing after an income increase.
The fraud dimension: when stacking is deliberate
Beyond financial distress, a subset of stacking behaviour is fraud. Ocrolus data from the US micro-lending market indicates that a borrower who applies for a second loan within 15 days of a first is approximately 4x more likely to be engaged in deliberate stacking behaviour; a third application within 15 days carries approximately 10x the fraud risk indicator. While this is US data, the underlying pattern (rapid multi-lender borrowing as a fraud indicator) is plausible in SA micro-lending given similar compressed application timelines and multi-lender borrower behaviour.
Deliberate fraud stacking has a slightly different bank statement profile than distress stacking. In fraud cases, the disbursements are often withdrawn as cash within hours of receipt rather than being used to service existing obligations. The account balance does not build up; each disbursement flows straight out. This cash-out pattern, combined with rapid application timing, is the fraud-specific signature that bank statement analysis can surface before the bureau reflects the accumulating obligations.
The SA market context amplifies this risk. With 18.08 million credit applications in Q1 2025 alone, and a sector that includes thousands of NCR-registered micro-lenders at varying levels of process maturity, the structural opportunity for stacking is significant. The bureau lag creates a window; distress and fraud both exploit that window.
How to set up detection that doesn't rely on bureau timing
Detection requires obtaining and analysing a three-month bank statement at origination. Bureau-only assessment leaves the stacking window permanently open. Once the statement is in hand, the analysis looks for the pattern combination described above: new lender income credits, new financial-services debit orders, and cash withdrawal changes, all within a compressed time period relative to the most recent month of the statement.
For lenders processing volume, manual review cannot sustain this. A trained analyst reviewing a bank statement manually takes an estimated 30 to 45 minutes to categorise transactions and identify stacking patterns. At any meaningful application volume, the review becomes a bottleneck or is abbreviated to the point where stacking signals are missed.
AffyScore's statement analysis detects loan stacking through automated transaction categorisation: multi-lender income credits are flagged as a stacking signal, multiple new debit orders to financial-services payees are identified, and cash withdrawal pattern changes are tracked against the three-month baseline. These patterns appear in the statement before the bureau reflects the obligations. In testing, the combination of income credit categorisation and debit order payee analysis surfaces stacking indicators that bureau-only assessment structurally cannot capture in the relevant window.
The output feeds directly into the obligation enumeration that short-term lenders need for a complete Reg 23A affordability assessment. Stacking detection is not a separate product or workflow. It is the same bank statement analysis that is structured to address the “all monthly obligations” verification requirement, applied to the signals that indicate obligations taken out this month rather than obligations listed on bureau.
