Every NCR-registered micro-lender in South Africa must conduct an affordability assessment before granting credit. NCA Section 81 requires it. There are no shortcuts. Before a rand leaves the building, the lender must prove the borrower can afford the repayment.
Today, that proof is assembled by hand. A human analyst opens a bank statement PDF, identifies income sources, categorises expenses into the NCA's prescribed categories, checks for red flags (gambling transactions, lender stacking, returned debits), calculates disposable income per Regulation 23A, and writes up the result. For a single file, this takes 20 to 45 minutes. Two analysts reviewing the same statement routinely produce different results. There is no standardised audit trail. And when the NCR investigates a reckless lending complaint, the lender's compliance evidence is a spreadsheet someone built in-house.
About 16 million South African adults sit outside the formal credit system. No bureau history, no score. But they all have bank accounts. The bureau tells you what happened in the past. The bank statement tells you what's happening right now.
Several South African companies extract transaction data from bank statements. OCR, open banking APIs, screen scraping. That problem has five or six players, and they do it competently. But extraction is step one of a five-step process.
After you extract transactions, you still need to categorise them into 17 NCA expense categories, check the document for tampering, calculate Reg 23A affordability with an audit trail, identify behavioural risk signals, and recommend the best collection day. Existing tools hand back structured data. The analyst still has to do the work. It's faster than reading the raw PDF, but the 45-minute review becomes a 25-minute review, not a 1-minute automated decision pack.
The gap isn't technical. It's that nobody has packaged the full back-office workflow as one product for the credit industry.
Advisory only. A trust signal alongside the decision, never part of it.
Built for SA's major banks. Scanned, photographed, or native PDF, it handles all of them. Regex-first extraction with AI vision fallback. Processing time: under 60 seconds.
NCA Section 81 is clear: the credit provider bears reckless lending liability. Not the tool vendor. Not the bureau. The lender.
Tools that claim to "automate lending decisions" or "auto-approve" put their customers at regulatory risk. If the NCR investigates a reckless lending complaint and finds the lender delegated the affordability assessment to a black-box algorithm with no human oversight, the lender has a problem.
Affy's design is deliberate. The output is Clear / Caution / Review, never Approve or Decline. The tamper checks are advisory: they flag, they never block. Every risk driver comes with a plain-English explanation the credit officer can read, verify, and act on.
This isn't a limitation. It's why compliance-conscious lenders will choose Affy over tools that promise to remove the human from the loop. The human stays in the loop because the law requires it. Affy makes that human faster and more consistent, with a complete audit trail.
The addressable market starts with roughly 4,500 NCR-registered micro-lenders and about 2,200 debt counsellors. That's 6,700 entities, most doing affordability assessments manually or not at all.
CASA (Credit Association of South Africa, formerly MicroFinance SA) represents 1,800+ non-bank credit providers. Four CASA-listed administration software companies, ACPAS, Mycomax, Fintech Fundi, and Altron FinTech, serve micro-lenders but have no bank statement extraction or affordability automation. They are potential customers, not competitors.
Affy's pricing is transparent: R8 to R35 per extraction, published on the website. Every competitor hides pricing behind "Contact Sales". No competitor ranks for Reg 23A affordability keywords. No competitor runs Google Ads in this category. The paid and organic search landscape is wide open.
The blog pipeline (39 posts planned, 4 live) is building search authority for a category no competitor has claimed. Early traction: 4 active prospects across micro-lending and financial services, including one design partner providing real operator requirements.
The brand is Affy. The first product is AffyScore. But the score isn't the headline value, the work is. When we describe what Affy does, we describe what the analyst does today, then show that Affy does it in one minute.
The positioning one-liner: "What your credit analyst does in 45 minutes, Affy does in one."
Extraction is table stakes: 5+ companies do it. We lead with what happens after extraction: categorisation, affordability, tamper checks, risk drivers, collection day. That's the work nobody else packages.
For context. Competitors are mentioned where relevant throughout this document. This appendix provides the full landscape for reference, including companies surfaced from the CASA service providers directory that were not in our original research.
Extraction + categorisation. Stops before affordability calc. Cape Town, ~16 staff. No Google Ads. Declining SEO (76 to 44 ranked keywords, Mar-Jun 2026).
Enterprise verification platform. VLM-based extraction (R29-R219/statement). Zero organic traffic. Brand declining in search.
SME alternative credit scoring. Bank statement OCR via Stitch. Generic 0-100 score. $1.1M pre-seed. SME focus, not consumer micro-lending. No Reg 23A. From CASA scan.
OCR + KYC + biometrics + Mifos core banking. Pretoria, 2-10 staff. Binary eligibility output, no Reg 23A affordability. From CASA scan.
Holds SERP #1 for "bank statement analysis SA". IT services company with a fragile organic position. Not a product company.
Broad data orchestration (identity + credit + AML + property). Bank statement parsing as a feature, not the focus. Enterprise/property market. VC-backed, 17 staff. From CASA scan.
Open banking API. Watch list: existential if SA open banking regulation arrives and mandates API-based data sharing.
Bureau integration for debt counsellors. No bank statement extraction capability.
Credit industry administration software. Based in Centurion. No bank statement extraction.
Cloud micro-lending administration platform. Pretoria-based, industry leader in LMS. No extraction capability.
SaaS loan origination and management system for small microfinanciers. No extraction.
Webfin/Delfin ecosystem. Bureau-based affordability only, no bank statement analysis.
The brand is Affy. The domain is affy.co.za. That's settled. The question is what the product is called. Right now it's AffyScore, which anchors every conversation on a number. If the value is the work, the name should point at the work.
Six options, with thinking on each:
Keep what we have
Established with prospects. Domain matches. But it anchors on the score, which we've agreed isn't the lead value. Every time someone hears "AffyScore" they think credit score, not decision pack. Works if we reframe "Score" internally to mean "the full assessment" but that's a stretch.
Risk: keeps the conversation on the number, not the work.
Points directly at what we do
An affordability assessment is the legal requirement (NCA Section 81). The name says exactly what the product delivers. "Affy" for short, "AffyAssess" when you need the full name. Lenders, compliance officers, and NCR auditors all understand "assessment" instantly. No explanation needed.
Strongest alignment with work-centric positioning and regulatory language.
The preparation before the decision
Implies the product prepares the decision pack, not makes the decision. Reinforces the "human decides" posture. Short, punchy, easy to say. Downside: "prep" is vague. Doesn't tell you what's being prepared. A compliance officer hearing "AffyPrep" for the first time wouldn't know it's about affordability.
Clean but requires explanation.
Matches the regulatory language we already use
We renamed "Verdict" to "Recommendation" to "Signal" in the product. The output is a signal, not a decision. AffySignal is consistent with that evolution. But "signal" is abstract. It works for people already in the conversation, less so for cold prospects. A Google Ad saying "AffySignal" tells you nothing.
Internally coherent but too abstract for market-facing.
Named after the output
The product delivers a "decision pack". AffyPack names the thing the lender actually receives. Concrete, memorable, slightly informal. "Send me the AffyPack" works in conversation. Downside: could sound like a bundle or package deal rather than a technology product. Less gravitas than AffyAssess.
Memorable but might feel lightweight for regulated buyers.
Drop the suffix entirely
One word. The brand IS the product. Like Stripe, or Plaid. Simple, clean, domain-ready. "Run it through Affy" is the shortest possible sentence. Downside: if Affy later launches a second product, you need a naming convention. But that's a future problem. For now, one product = one name works.
Simplest option. Defer the naming problem until there are two products.
AffyAssess is the strongest fit for where the positioning is going. It says what the product does in the language the buyer already uses. "Assessment" is the word in the NCA, the word the NCR uses, the word every compliance officer types into their audit report. It's not clever, it's clear. And clear wins in B2B.
Second choice: just "Affy". Skip the suffix, let the product speak for itself. Use "Affy" in all external copy now, and decide on a formal product name after Welltec and Raoul have used it for a month. They'll tell you what they call it.
Either way, stop leading with "Score" in new materials. The score is an output, not the product.