# ExactRec

> ExactRec is automated cash application for asset finance. It matches incoming bank payments to open receivables — specifically the residual that rules-based matching inside a lessor's ERP cannot handle, which typically plateaus somewhere between 70% and 80%. Every transaction lands in one of three buckets (auto-applied, suggested, unmatched), and ExactRec produces an allocation file in whatever format the client's system accepts for import.

ExactRec is part of Titan Fintech, built by the team behind ExactSum (AI bank statement analysis), ExactCov (covenant monitoring) and ExactVal (equipment valuation). v1 is deliberately file-in, file-out: no ERP integration, because write-back is where cash application projects stall and it is not needed to prove the value.

## Why asset finance is the hard case

A single payer covers several contracts in one lump sum, part-pays, pays late, sends no remittance advice, quotes the wrong reference, or bundles rentals with VAT, insurance and maintenance. When a direct debit fails, arrears allocation rules compound the problem. The matching is fuzzy, contextual and text-heavy — the shape of problem where a model materially outperforms rules — and the output is verifiable: a proposed match either reconciles to the penny or it does not.

## What the engine does

- Exact reference matching — the cheap wins, establishing parity with the client's existing ERP rules
- Fuzzy payer-name matching against the customer master: trading names, abbreviations, group entities paying for subsidiaries, personal names on sole-trader accounts
- Lump-sum splitting — one payment allocated across several contracts; the differentiator in asset finance and the point at which rules-based systems fail
- Part-payments and overpayments, with a proposed allocation
- Direct debit batch reconciliation and identification of failed collections

## Inputs

Three files, all of which a client can typically export within a day:

- Bank statement transactions (parsed by the existing ExactSum pipeline)
- Open items ledger — unpaid invoices and rentals with contract reference, customer, due date and amount
- Customer master — names, account references and any known payer aliases

## Output

- Auto-applied — above the client's confidence threshold; reference, payer and amount all reconcile. No action required.
- Suggested — a ranked proposal with the matching logic shown, including proposed splits across contracts. One-click accept or reject.
- Unmatched — no credible candidate found; reviewed manually, but with a stated reason rather than a blank.

Plus an allocation file in the client's import format. The confidence threshold separating auto-applied from suggested is client-controlled.

## Out of scope for v1 (deliberately)

- ERP write-back and live system integration
- Collections, dunning and customer communication
- Multi-currency handling
- Learning from user corrections — the v2 capability, deferred because the correction data does not exist yet

## The measure of success

A single number: auto-match rate against the client's current system, measured on the same month. ExactRec has no published match rate yet; the first pilots will produce one on real client data. The gap between the two figures is the entire commercial proposition, expressed in clerk-hours recovered and unapplied cash cleared.

## Pages

- [Home](https://exactrec.com/): the problem, the engine, the three output buckets and what v1 deliberately excludes
- [Run a pilot](https://exactrec.com/pilot): one month of three anonymised files in, one measured match rate out
- [Security](https://exactrec.com/security): security and compliance practices, zero data retention
- [Contact](https://exactrec.com/contact): contact the team

## Contact

- Email: hello@exactrec.com
- Demo bookings: https://exactrec.com/book-a-demo