One-to-many matching is when one bank-statement line clears several ledger lines, or one journal absorbs several bank movements. VLOOKUP and one-to-one matching fail here. Teams that force those formulas pile exceptions, then close the books with a guess.
The patterns you will see
An MSME loan payoff: one customer transfer covers principal, interest, and fees. A payment-gateway settlement: one inbound batch covers hundreds of orders. Payroll: one bank debit covers many salary journals. All of that is legitimate. What is not legitimate is dumping the remainder into “other difference” with no rule.
QRIS and BI-FAST make the pattern larger. Bank Indonesia recorded 12.55 billion QRIS transactions in the first half of 2026 worth about Rp 600.7 trillion, per Kompas on 18 September 2026. At that volume, settlement is almost always one-to-many. Do not design matching as if every order has one tidy bank line.
A safe rule order
First, unique references if they exist: virtual account, batch id, payoff code. Second, amount plus a written date window, for example D−1 to D+2. Third, description similarity only as a hint, never as the decider. Fourth, leftovers that are not unique go to a human queue.
Do not insert rows until the amounts fit. That is fabrication, not reconciliation. If a Rp 10 million credit clears three Rp 3 million invoices and leaves Rp 1 million, leave Rp 1 million as a named exception, not “rounding.”
Many-to-one and many-to-many
Many small movements clearing one invoice also appear, especially when customers drip-pay. Cap depth: at most N movements in M days for one invoice. Without a cap, the engine will glue unrelated transactions because the sums happen to match.
PBI 10/2025 and PADG 32/2025 took effect on 31 March 2026. Payment gateway sits in the forwarding bundle. An acquirer batch must still explode to orders, or an auditor will see one large number with no children.
Reconciliation and month-end close notes for Indonesian finance teams.
Every Monday, five minutes.
Trail for ICoFR
Store the match group: members, winning rule, who approved if a human stepped in. A spreadsheet that merges rows by copy-paste has no version. OJK’s IASC, as reported by Kompas on 8 September 2026, blocked Rp 771.2 billion tied to fraud. If one member of a group is frozen, do not silently split the group. Mark a “held” exception.
OJK Regulation 34/2025 on rural-bank IT takes effect on 18 December 2026. One-to-many matching that lives only on a laptop cannot be restored. Keep it in a system that is backed up.
National holidays that delay QRIS settlement, as in Idul Adha 2026 merchant notices, shift batch dates. The matching window must read the settlement calendar, or every long weekend becomes a pile of wrong pairs.
Frequently asked questions
What is one-to-many matching?
One bank movement clearing several journals, or the reverse. Normal in payoffs, payroll, and gateway settlement.
Why does Excel fail?
Common formulas look for one pair. Remainders and combinations are handled by hand with no trail.
Must amounts always fit exactly?
No. An honest remainder is better than a fictitious line that makes the total work.
How do we limit false merges?
Reference first, then amount and date window, then a depth cap. Description is only a hint.
Do we still need people?
Yes, for ambiguous leftovers and frozen accounts. The engine does the unique cases.
What evidence should we keep?
Group members, rule, timestamp, and human approval if any. That is what auditors ask for.
Test Threadion on one month of your own bank data
Threadion
