Cash Application That Doesn't Stop at 80%.
Rules-based matching inside your ERP plateaus somewhere between 70% and 80%. The residual gets worked by hand every morning — while unapplied cash sits on the balance sheet, credit limits stay blocked and collections chase customers who have already paid. ExactRec works that residual.
Incoming payment · unmatched by ERP rules
NORTHGATE PLANT LTD
Matched to Northgate Plant Hire Limited in the customer master — trading-name variant, no remittance advice received
Amount received
£47,382.60
Proposed allocation across 3 open items
AF-2024-8817
Monthly rental · due 01 Aug
£18,240.00
AF-2023-6104
Monthly rental · due 01 Aug
£22,918.60
AF-2025-1190
Maintenance · due 15 Jul (arrears)
£6,224.00
Allocated
£47,382.60
Residual
£0.00
Confidence 96% · reconciles to the penny
Suggested — one-click accept
Illustrative example. ExactRec proposes the split; nothing is applied without your confirmation.
Built by the team behind ExactSum — AI document intelligence trusted by lenders and audit firms
70–80%
where rules-based ERP matching typically plateaus
3 files
everything ExactRec needs, all exportable in a day
Zero
ERP integration required to run a pilot
1 number
your match rate, measured against your current system
Why asset finance is the hard case
Cash application is the matching of incoming bank payments to open receivables. In asset finance it is unusually hard — and the ways it breaks are exactly the ways rules cannot bend.
One payment, several contracts
A single payer covers several agreements in one lump sum. Splitting that correctly across open items is where rules-based systems fail — and it is the everyday case, not the edge case.
No remittance advice
The payment arrives with nothing to say what it is for. The reference field holds a customer name, an old contract number, or nothing usable at all.
The wrong reference
Customers quote a settled agreement, a quote number, or the reference from a contract refinanced two years ago. The characters match nothing in the ledger.
Bundled components
Rentals arrive bundled with VAT, insurance and maintenance. The gross figure ties to no single open item, so an exact-amount rule never fires.
Part-payments and overpayments
The customer pays what they can, or rounds up. A near-miss on amount is treated as no match at all, and the whole receipt drops to the manual queue.
Failed direct debits
When a collection fails, arrears allocation rules compound the problem — and reconciling the batch back to individual mandates becomes its own morning job.
This residual is what ExactRec addresses. 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.
Export, Match, Review, Apply
Export
Send three files your system already produces: bank statement transactions, the open items ledger and the customer master. No integration, no IT project.
Match
ExactRec runs exact-reference, fuzzy payer and lump-sum splitting logic across every unmatched receipt, scoring each candidate allocation.
Review
Results land in three buckets. Auto-applied needs nothing from you; suggested allocations are accepted or rejected in one click; unmatched items carry a stated reason.
Apply
ExactRec writes an allocation file in whatever format your system takes for import. You stay in control of what gets posted.
Three buckets, and never a guess dressed as a certainty
Every transaction lands in exactly one bucket. The three-bucket structure is what makes the product trustworthy: ExactRec never asserts a confidence it does not have, and the threshold that separates auto-applied from suggested is a dial you control.
| Bucket | What it means | Action required |
|---|---|---|
|
Auto-applied
|
Above your confidence threshold. Reference, payer and amount all reconcile. | None — allocated straight to the output file. |
|
Suggested
|
A ranked proposal with the matching logic shown — including proposed splits across contracts. | One-click accept or reject by the AR team. |
|
Unmatched
|
No credible candidate found. | Reviewed manually — but with a stated reason, not a blank. |
Above your confidence threshold. Reference, payer and amount all reconcile.
Action required: None — allocated straight to the output file.
A ranked proposal with the matching logic shown — including proposed splits across contracts.
Action required: One-click accept or reject by the AR team.
No credible candidate found.
Action required: Reviewed manually — but with a stated reason, not a blank.
Alongside the buckets, ExactRec produces an allocation file in whatever format your system accepts for import.
Five passes over
every unmatched
receipt.
ExactRec layers cheap, certain matching first and escalates to the harder logic only where it is needed — so the easy wins stay cheap and the difficult residual gets the attention.
Exact reference matching
The cheap wins first, establishing parity with the rules your ERP already runs. If your system would have matched it, ExactRec matches it too — so the comparison starts from level ground.
Fuzzy payer-name matching
Payer names are matched against the customer master across trading names, abbreviations, group entities paying for subsidiaries, and personal names on sole-trader accounts.
Lump-sum splitting
One payment allocated across several contracts. This is the differentiator in asset finance and the precise point at which rules-based systems fail — a proposed split either reconciles to the penny or it is not offered.
Part-payments and overpayments
Where the amount does not tie exactly, ExactRec proposes an allocation rather than dropping the whole receipt into the manual queue — with the shortfall or excess stated explicitly.
Direct debit batch reconciliation
Collection batches are reconciled back to the underlying mandates, and failed collections are identified as failures rather than left as unexplained gaps in the ledger.
Three files. Nothing else.
Every input is something your finance system already produces, and a client can typically export all three within a day.
Bank statement transactions
Incoming receipts as they arrive from the bank — parsed by the existing ExactSum pipeline, in whatever format your bank provides.
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 — the ground truth that fuzzy payer matching is scored against.
Because v1 is file-in, file-out, the platform you run on does not have to support anything in particular. If it can export a ledger, it can run a pilot.
One cleared residual, four teams that feel it
Accounts Receivable
The residual is worked by hand, every morning, by the accounts receivable team. It is skilled, repetitive work: reading a payer name, working out which entity it belongs to, then splitting a lump sum across open items until the arithmetic closes.
ExactRec does the first pass. Receipts that reconcile cleanly are applied; the ambiguous ones arrive as ranked proposals with the matching logic shown, so the decision takes a click rather than a spreadsheet. Nothing is posted without your confirmation.
The team stops starting each day behind, and spends its judgement on the genuinely hard cases instead of the mechanical ones.
Unapplied cash is rarely just an accounts receivable problem. It blocks credit lines, misdirects collections and distorts the month-end position — so clearing it pays out across the business.
A single number decides this
Not a feature list, and not a demo on our data. The auto-match rate ExactRec achieves on your book, measured against what your current system achieves on the same month.
72%
Your ERP today
89%
With ExactRec
17 pts
The gap
These figures illustrate how the comparison is expressed; they are not a result. ExactRec has no published match rate yet — the first pilots will produce one, on real client data. Whatever that gap turns out to be is the entire commercial proposition: clerk-hours recovered and unapplied cash cleared.
"Cash application is the rare finance problem where the answer is checkable. A proposed allocation either reconciles to the penny or it does not, so nobody has to take a model's word for anything. That is also why we will not lead with a match rate we have not measured — one month of a real client's data is worth more than any amount of further specification"
Aleks Kudic
Co-founder, ExactRec — previously CTO at Funding Options
Built for the Residual, Not the Easy 80%
ExactRec is aimed squarely at the receipts your current rules leave behind — and at making every proposal it puts in front of you checkable in seconds.
Matching Capabilities
Exact Reference Matching
Establishes parity with your existing ERP rules, so the comparison between the two starts from level ground
Fuzzy Payer Matching
Trading names, abbreviations, group entities paying for subsidiaries and personal names on sole-trader accounts, scored against the customer master
Lump-Sum Splitting
One receipt allocated across several contracts — the asset finance differentiator, and the point where rules-based matching gives up
Part & Overpayments
A proposed allocation with the shortfall or excess stated, instead of the whole receipt dropping into the manual queue
Direct Debit Reconciliation
Collection batches reconciled to the underlying mandates, with failed collections identified as failures
Bundled Component Handling
Rentals arriving bundled with VAT, insurance and maintenance are decomposed against the open items they settle
Confidence Threshold Dial
You set the line between auto-applied and suggested. Tighten it while you build trust, loosen it once the numbers hold up
Shown Matching Logic
Every suggestion states why it was proposed — which payer alias, which open items, which arithmetic — so a reviewer can check it in seconds
Stated Reasons on Unmatched
An unmatched receipt arrives with a reason rather than a blank, so manual review starts from a lead instead of from scratch
Working With the Output
Allocation File Export
Output in whatever format your system accepts for import — the whole reason a pilot needs no integration work.
Full Audit Trail
Every allocation traces back to the receipt, the matching logic and the open items it settled, with reviewer and timestamp on anything accepted.
Runs on Exported Files
No agent to install, no database connection, no change control. Three exports in, one allocation file out.
What v1 does not do — on purpose
Cash application projects stall on scope. These are deliberate exclusions from the first version, not gaps we have overlooked.
ERP write-back and live integration
Write-back into your ERP or ledger system is where these projects stall, and it is not needed to prove the value. ExactRec outputs a file your system imports. Integration follows once the match rate is established.
Collections, dunning and customer communication
ExactRec clears the residual so your collections process works from a true list. It does not contact your customers.
Multi-currency handling
v1 works a single currency. Multi-currency books are on the roadmap, not in the first release.
Learning from user corrections
Every accept and reject is training signal, and using it is what makes the product defensible over time. It is deferred deliberately: that correction data does not exist yet, and it will come from the first pilots.
Industry leading security
Cyber Essentials Certified
ExactRec security practices and controls have been independently audited by experts.
Encryption
Data is encrypted in transit and at rest using the most secure algorithms available.
Zero Data Retention
Your files are never used to train AI models and are deleted permanently once you remove them.
Send one month. Get a measured match rate.
The ask is narrow and costs you almost nothing: one month of the three input files, anonymized. From that we report a measured match rate against your current system, and a worked example of lump-sum allocation on your own data.