You run the affordability assessment. The consumer qualifies. Income checks out, obligations are within ratio, and the Reg 23A numbers look clean. You set the debit order for the 1st, press go, and wait for confirmation. The debit bounces.
What went wrong?
You asked the right question (can they afford it?) and got a truthful answer. But you missed the second question, the one that determines whether the money is actually there on the day you try to collect it: when should we collect?
This is the collection date signal. It is the gap between affordability and collectability, and in AffyScore's testing across a limited sample of micro-lending portfolios, it appears to account for a material share of first-instalment failures that have nothing to do with whether the consumer can afford the repayment.
The timing problem hiding in plain sight
Most lenders default to one of two debit order dates: the 1st or the 25th. Some offer the consumer a choice between a handful of fixed dates. The logic is administrative convenience; batch processing is simpler when everyone debits on the same day.
Salary dates in South Africa vary significantly. Government employees are commonly paid mid-month, while many private-sector workers receive salary around the 25th or the last business day of the month, though dates vary by employer and department. Construction, hospitality, and gig workers are often paid weekly or fortnightly, with no consistent calendar date.
When a debit order runs on the wrong day, it competes with every other obligation hitting the account that date. Rent. Groceries. Insurance. Other debit orders also set for the 1st, because their lenders also chose the default. The result is a stack of competing claims on an account that may have already been drawn down to near-zero.
The consumer can afford your instalment. They proved it on paper. But the money was there on the 26th, not the 1st, and by the 1st it was gone.
What the collection date signal actually is
AffyScore analyses three months of bank statement balance patterns and returns three data points that a standard affordability check does not provide:
- Best debit day: the specific day of the month when the account is most likely to have sufficient cleared funds. This factors in recurring inflows, the timing and size of outgoing obligations, and the balance trajectory across the full statement period, not just salary date alone.
- Collectability %: the proportion of observed months during which the account held adequate funds on the recommended day. A collectability score of 92% means the account had enough on that day in roughly nine out of ten observed months. A score of 55% tells a different story.
- Predictability %: how consistent the balance pattern is month to month. A salaried employee with a fixed pay date and stable obligations will show high predictability. A gig worker with irregular deposits and variable expenses will show low predictability, not necessarily because they cannot afford the repayment, but because the optimal collection day shifts.
These three signals turn the collection decision from a guess into a data-driven recommendation. The affordability assessment tells you whether to lend. The collection date signal tells you when to collect.
How it works under the hood
AffyScore uses what we call balance curve analysis. The system maps the daily closing balance across the full statement period (typically 90 days) and overlays three pattern layers.
First, income identification. The system detects salary deposits, grant payments, and other recurring inflows by amount, frequency, and source. It distinguishes a monthly salary from a weekly wage, a child support grant from a freelance invoice, and a once-off payment from a recurring one.
Second, obligation mapping. Recurring debit orders, EFT payments, and cash withdrawals that follow a monthly pattern are identified and plotted against the balance curve. This reveals when the account is under the most pressure, typically the first week of the month, when rent, insurance, and other debit orders cluster together.
Third, trough and peak detection. The balance curve shows a predictable shape for most salaried consumers: a peak shortly after payday, a steep decline as fixed obligations debit, a trough in the days before the next salary, then the cycle repeats. The best collection day sits in the window between income arrival and the onset of major outflows, after the money lands but before it leaves.
This analysis runs automatically as part of the standard bank statement scoring process. No additional documents. No extra API calls. The collection date signal is included in the same decision pack that delivers affordability, income verification, and the behavioural credit score.
The cost of getting collection timing wrong
Bounced debit orders are not free. Based on publicly available bank fee schedules, the bank penalty alone for a failed DebiCheck collection ranges from roughly R50 to R80. Add re-presentation charges and the administrative overhead of tracking and retrying, and the total cost per failed attempt can reach R100 to R150. Actual costs vary by bank, processing channel, and internal operations.
To illustrate: a lender processing 1,000 collections per month with a hypothetical 10% bounce rate absorbs 100 failed debits. At an illustrative average cost of R75 per bounce, that is R7,500 per month (R90,000 per year). Scale to 10,000 collections and the figure becomes R900,000 annually. These are illustrative estimates, not empirical AffyScore outcomes, but they show how quickly bounce costs compound. This also excludes the downstream cost of the collections cascade that a bounced first instalment can trigger.
That cascade is where the real damage sits. A bounced first instalment is one of the stronger predictors of eventual default in short-term lending portfolios, not because the consumer cannot afford the loan, but because the failed collection triggers escalating consequences. Late fees compound. The following month now carries the missed payment plus the current one. Subsequent debit orders are more likely to fail because the account never fully recovers. What started as a timing problem becomes a credit problem.
In AffyScore's testing across a limited sample of micro-lending portfolios, first-instalment failures appear more frequently linked to collection timing than to affordability failure alone. The consumer qualified. They could afford it. The money was not there on the day it was collected.
DebiCheck and PASA's mandate
The Payment Association of South Africa (PASA), operating under the National Payment System Act overseen by the South African Reserve Bank, has progressively rolled out DebiCheck across the banking system. Under DebiCheck, consumers must authenticate each debit order mandate before it can be processed, a significant improvement over the older system, where unauthorised debits were a persistent problem. The National Credit Regulator (NCR) has separately reinforced, through its compliance monitoring of affordability assessments, that lending decisions should account for the consumer's demonstrated capacity to repay.
Authentication is not the same as availability. A consumer can authenticate a debit order for R2,500 on the 1st of every month and still have insufficient funds when the collection runs. DebiCheck guarantees the consumer agreed to the debit. It does not guarantee the money will be there.
This is precisely where the collection date signal adds value. It tells you not just that the consumer has authenticated a mandate, but when to schedule that mandate for the highest probability of clearing. You are not choosing between the 1st and the 25th based on administrative preference. You are choosing based on data that shows when the account is most likely to have funds.
For lenders already implementing DebiCheck workflows, the collection date signal slots directly into the mandate setup process. Authenticate the debit order for the recommended day, not the default.
A concrete example
To illustrate with a hypothetical consumer earning R22,000 per month, paid on the 25th. Three months of statements show a consistent pattern:
- 25th–27th: Balance peaks at R18,000–R20,000 after salary clears and before major debits run.
- 28th–1st: Rent (R8,500), car insurance (R1,200), and two existing loan repayments (R1,800 combined) debit. Balance drops to R6,500–R8,000.
- 2nd–7th: Groceries, fuel, and ad-hoc spending bring the balance to R2,000–R3,500.
- 8th–24th: Balance drifts between R800 and R2,000, occasionally dipping below R500.
A collection set for the 1st competes with rent and insurance on a rapidly declining balance. If rent debits first (which it will, authenticated earlier), your R2,500 instalment is fighting over whatever remains.
Set the collection for the 26th and you are first in line after payday. The account is at its peak. Your R2,500 debits from an R18,000 balance instead of a R6,500 balance. Same consumer. Same income. Same affordability outcome. Entirely different collection probability.
Based on a pattern this consistent, AffyScore's collection date signal would be expected to flag the 26th as the best debit day, return a collectability score above 90%, and note high predictability given the stable salary date and obligation pattern. The lender does not need to work this out manually. The signal is in the decision pack.
Who benefits most
The collection date signal is relevant to any credit provider that collects via debit order, but it delivers the most value in portfolios where:
- The borrower profile is lower-to-middle income. These consumers carry less buffer between income and obligations. Timing matters more when there is no surplus sitting in the account.
- Collection volumes are high. Micro-lenders, short-term lenders, and telcos running thousands of debit orders per month see the bounce-rate reduction compound quickly.
- Income is irregular. Gig workers, commission earners, and seasonal employees show variable balance patterns. The default collection date is often wrong for these consumers, and the predictability score flags exactly how much uncertainty exists.
- First-instalment default is a known problem. If your portfolio data shows a disproportionate number of defaults beginning with a bounced first collection, collection timing is likely a contributing variable.
Affordability answers one question. Collectability answers another.
The lending industry has spent years refining affordability assessment. Income verification has improved. Obligation detection is more accurate. Bureau data provides a longer-term repayment history. But affordability is a necessary condition for successful lending, not a sufficient one.
A consumer who can afford a repayment but whose account is empty on collection day is not a credit risk in the traditional sense; they are a timing risk. And timing risk is solvable with data that already exists in the bank statement.
The collection date signal does not replace the affordability assessment. It completes it. It answers the question that affordability leaves open: not whether to collect, but when.
To our knowledge, collection timing is not a standard output in most bank statement decision packs in South Africa. It is a standard output in every AffyScore decision pack.
