Why differences appear at all
- Timing: the sub-ledger posts on trade date, the GL on settlement, and the difference is legitimate until it is not.
- Mapping: a product or fee type posts to an account the mapping did not anticipate.
- Manual journals: an adjustment in the GL with no corresponding sub-ledger movement.
- Failed interfaces: a batch that partially posted and was never fully replayed.
- Rounding and FX revaluation, which are small individually and structural in aggregate.
Suspense accounts deserve a workflow, not a balance
A suspense balance is not a number, it is a set of items each of which somebody must resolve. When suspense is reconciled item by item, with an owner and an age per item, the balance falls because the work is visible. When it is reviewed as a balance, it grows, and the review becomes an explanation of the growth.
Input
- 14 items, 2,318,440.00 USD total
- Oldest item 63 days, no owner recorded
- Nine items share the same fee type
Output
- Nine items grouped as Fee not accrued, one accrual fix, one approval.
- Three items Timing, cleared by the next settlement cycle, tolerance proposed and approved.
- Two items aged past threshold, escalated with an owner and a due date.
Mapping is the control everybody forgets
Sub-ledger to general ledger reconciliation depends on a mapping: which sub-ledger movement belongs to which GL account. That mapping is usually built once during an implementation and then changed piecemeal as products launch, fees change and entities reorganise. Because a mapping change rarely breaks anything visibly, it accumulates without review, and a surprising proportion of persistent GL differences trace back to a mapping decision nobody can now explain.
- Every mapping rule should have a version, an author and an effective date, exactly like a matching tolerance.
- A new product or fee type should require a mapping decision before it can post, rather than defaulting to a suspense account.
- Unmapped movements should raise an exception with an owner, not accumulate silently.
- The mapping should be readable by the finance team without a developer translating it.
The four sources of a GL difference
| Cause | What it looks like | Correct handling |
|---|---|---|
| Timing | Sub-ledger posts on one date, GL on another | Expected item under an approved window, cleared automatically |
| Mapping | A movement posted to the wrong account | Mapping correction, versioned, with the population restated |
| Manual journal | A GL entry with no sub-ledger movement behind it | Item with an owner and supporting evidence attached |
| Genuine error | A posting that should not exist | Reversal under maker-checker with a reason code |
Suspense accounts, and why they quietly grow
A suspense account exists so that a posting with an unknown destination does not stop the day. That is a sensible mechanism and it becomes a problem for one reason: nothing in the process makes the balance uncomfortable. An item parked in suspense has no owner, no age that anyone reports on, and no requirement to be resolved before the next one arrives, so the population grows by small amounts until it is large enough to be a finding.
Treating suspense as a reconciliation in its own right fixes it without any new policy. Every item that enters carries the reason it could not be posted, the person who parked it and the date the clock started. Every item that leaves carries the account it went to and who approved the move. The balance is then not a number to be explained at quarter end but a queue with an age profile, and a queue with a visible age profile tends to get worked.
What a reconciled ledger is worth to the people who never open it
The finance team feels the close. Everyone else feels the consequences of an unreconciled ledger without connecting them to it: a business line reading a margin that later restates, a treasury desk funding against a balance it does not fully trust, a regulatory return assembled from figures that needed a footnote. None of those are accounting problems in the moment they are felt, which is why the reconciliation is usually funded as a control cost rather than as what it actually is, which is the point where the bank decides whether its own numbers are usable.
Only the last two need a person. The first two are configuration problems dressed as daily work, and a reconciliation that classifies them correctly makes that obvious within a month rather than within a year.
Manual journals, and the reason they attract attention
A manual journal is a posting made by a person rather than produced by a process, and it is the entry an auditor looks at first because it is the entry with the fewest built-in controls. The reconciliation view of a manual journal is simple: it is an unmatched GL movement, and it should carry the same requirements as any other open item. An owner, a reason, supporting evidence attached at the time, and an approver who is not the person who posted it. A reconciliation that treats manual journals as ordinary items removes an entire category of quarter end question.
Reconciling daily when the ledger closes monthly
A common objection is that the general ledger is a monthly instrument and reconciling it daily is over-engineering. In practice the opposite holds. Daily reconciliation of the sub-ledger to GL movement, rather than of the balance, catches a mapping error on the day it starts instead of after it has produced thirty days of misposted entries. The monthly balance reconciliation then confirms a position that is already known to be right, which is why banks that move to daily movement matching usually report that the month end became uneventful rather than faster.
Certification without the scramble
An account is certified when its reconciliation is complete, its open items are classified and owned, and a reviewer who is not the preparer approves it. Because that state is a system fact rather than a signature on a printout, the evidence for the period is an export rather than a project. The finance framing of the same engine is on financial reconciliation software, and the sign-off view is balance sheet reconciliation.
Questions about general ledger reconciliation software
Can it read our chart of accounts?
The account structure comes from your GL export, and account level configuration such as tolerances, owners and certification cadence is applied on top of it. The chart of accounts stays owned by finance.
Do you support multi-entity and multi-currency?
Yes. Entity is part of the account model and currency is part of the item, with matching restricted within currency and revaluation differences classified rather than absorbed silently. Multi-entity is the Institution plan.
How are manual journals handled?
A journal with no sub-ledger counterpart is an unmatched item like any other: classified, owned, aged and evidenced. That is usually the fastest way to find out how many are being raised and why.
Can we reconcile intercompany balances?
Intercompany is a two sided reconciliation between entities and runs on the same engine, with the difference that both sides are yours and the classification vocabulary can be shared.