The matching engine
Six passes, in order, every time
Most reconciliation tools run one rule set and give up. Reconcile runs an ordered cascade — each pass only sees what the pass before it could not explain — so the easy 80% clears in milliseconds and the hard 20% arrives with its reasoning attached.
6
ordered passes, deterministic and replayable
What it does
Inside Reconcile
One-to-one, many-to-one, many-to-many
A bulk file that leaves the account as a single debit is matched back to every instruction inside it by bounded subset-sum, not by asking an operator to tick boxes.
Reference repair
Sending banks truncate, reformat and lose references. A sliding-window similarity pass recovers the link when the amount and direction still agree, and records exactly how confident it was.
Fee and spread awareness
Gross captures against net settlements clear inside a contractual band you define per acquirer. The deduction is posted to its own account rather than silently absorbed into revenue.
Cross-currency legs
Trade-time rates rarely equal fill rates. Conversion drift inside tolerance is matched and revalued; drift outside it becomes a break with the effective rate on the face of it.
Tolerances you can prove
Every threshold is a named, versioned parameter. Change one and the whole day re-runs, so you can see what a looser fee band would actually have let through before you approve it.
Nothing posts on a guess
Matches below your auto-post threshold are proposals. They queue for a human, they carry their evidence, and they never touch the general ledger unattended.
In depth
Why the order matters
Running every rule at once produces the worst of both worlds: false matches from loose rules, and no way to explain which rule fired. An ordered cascade means the strongest evidence always wins, and a weaker pass never gets the chance to claim a record that a stronger one would have taken.
- 01Pass 1 requires reference, amount, currency and value date to agree. Confidence 100%.
- 02Pass 2 keeps the amount fixed and repairs the reference. Confidence 90–98%, scored.
- 03Pass 3 allows a deduction inside the acquirer's contractual band and posts the residual.
- 04Pass 4 reconstructs batches by subset-sum, bounded to eight items and pruned on partial sums.
- 05Pass 5 converts foreign legs at the settlement rate and revalues the difference.
- 06Pass 6 proposes value-only pairings and always routes them to a human.
Specification
- Match types
- 1:1, N:1, 1:N, N:M
- Determinism
- Same inputs, same output, byte for byte
- Throughput
- 2.4M lines per run on standard infrastructure
- Tolerance controls
- Per source, per counterparty, per channel
- Replay
- Any historical run, at its original engine version
“The bulk payout problem is the one nobody solves. One debit on the statement, two hundred rows in our ledger, and every tool we trialled asked an operator to match them by hand. Watching the aggregation pass reconstruct a file from amounts alone was the moment we stopped evaluating vendors.”
Joris van Weel
Director of Payment Operations, Volta Mobility · Netherlands