I'm building a loan engine and I'd like your honest read.
I've spent the last while building fincalc, an engine for the maths every lender runs. Before I put more into it, I'm asking people I trust whether it's heading somewhere worth going. This takes about three minutes to read.
What it is
fincalc works out instalments, rates, what a customer can afford and the true cost of credit, all in one engine.
- It uses exact decimal arithmetic and a written rounding policy, so the same loan always gives the same figures.
- Fees and credit-life insurance are included in the APR, so the rate a customer sees reflects what the loan really costs.
- A compatibility mode can reproduce an older calculator's figures, so a lender can switch with no figure moving.
- When it can't do something yet, it says so and names the rule. It never gives a figure that only looks right.
A worked example
Here is the engine pricing a personal loan for a made-up lender, Acacia Credit:
- A customer with a R2 500 monthly budget can borrow up to R58 100.00.
- The APR is 30.82%, including fees and credit-life insurance, against a 22.75% nominal rate.
- On a simpler Acacia loan without fees, an introductory offer runs inside the schedule itself: 3 months at 1 934.62, then 2 284.07 for the rest of the term.
- 100 amount-and-term options come back in about a tenth of a second.
Where I think it could fit
It's strongest where someone has to recheck many existing loans and prove the answer. These are the five places I'm weighing. I want to know which ones are real.
Loan-book compliance checks
Recompute every loan in a book and compare it with the National Credit Act caps on interest and fees. The recomputation is in the engine today; the caps check would be built for this. The caps are tied to the repo rate, which has gone up twice in 2026 to 7.25%, so the check has to be repeated. There are 8,147 registered credit providers.
Sources: NCA Reg 42, eNCA on the September hike, NCR Annual Report 2024/25 (providers at 31 Mar 2025). Accessed 8 Oct 2026.
An engine for loan-software vendors
The smaller firms that sell loan software to SA lenders each build their own credit-act maths. fincalc could be that piece, kept current with every rate change.
Proving a system migration
When a lender changes core systems, show loan by loan that the new system gives the same instalments, interest and balances as the old one before switch-over, and explain any difference.
Checking customer refunds
The UK's car-finance redress scheme covers 12.1 million agreements and about £7.5bn of redress, within a £9.1bn total cost to firms. Parts of it are suspended until a tribunal hearing on 16–26 February 2027, so payments start in 2027 at the earliest. Redress figures will need checking, and re-running if the scheme changes. In Poland, the EU's top court upheld the rule that a wrongly stated loan cost can leave the borrower repaying capital only, and lawsuits have followed.
Sources: FCA statement, FCA PS26/3, MPA on the suspension, hearing dates (accessed 15 Oct 2026); CJEU C-472/23 (accessed 8 Oct 2026).
An independent check of the quoted rate
The EU's new consumer-credit rules apply from 20 November 2026, with one set formula for the annual rate shown to customers. fincalc could check that figure independently: the actuarial method behind it is already in the engine, and the EU's specific assumptions would be added.
Source: Directive (EU) 2023/2225. Accessed 8 Oct 2026.
What I'm asking you
- Gut check. Which of these sounds like a real need, and which sounds weakest? Is this worth building further at all?
- Introductions. Do you know anyone who works on any of them? Twenty minutes with them would help a lot. I'm learning, not selling.
- Circles. Where do these people talk shop: associations, forums, events, group chats?