Continuous Accounting: Closing the Books in Real Time
By Formance Staff
Accounting
Reconciliation
Embedded Finance
Loading article content
Loading table of contents
Loading author
Try Formance Ledger
Formance is open-source by design, so the ledger that enforces these properties is the same one you can read, run, and audit yourself. Clone Formance Ledger on GitHub, run it locally, and post a backdated settlement with distinct effective and recorded timestamps to see the as-of query for yourself.
For fintech and platform teams moving money across multiple payment rails, month-end is where the quarterly close breaks.
A settlement file from a payment provider arrives three days after month-end, and the total doesn't match your ledger. An engineer spends two days finding a five-figure gap in the fee logic, which sits in a service instead of the ledger. Finance holds the month open while the gap is explained, then reopens it when late chargebacks land. The close takes over a week.
Each step gets pushed to month-end because the ledger cannot be trusted between batches. If you move money across several payment rails, the ledger has to be trustworthy every day, with no spreadsheet reconstruction at month-end.
Continuous accounting is recording and matching transactions as they happen on a real-time ledger with automated reconciliation. The books are ready to close at any time.
How a ledger enforces continuous accounting on every write
Continuous accounting enforces ledger properties on every write. These are: events post exactly once, any period can be snapshotted at any timestamp, corrections use new postings, and external balances reconcile continuously with owned exceptions.
Formance is an open-source, programmable core ledger that unifies fiat and digital assets with regulatory-grade traceability. Formance Ledger and Formance Reconciliation enforce these properties on every write, so continuous accounting is the default behavior of the modules.
Every payment-rail event posts on write, exactly once
Every payment-rail event lands in the ledger as one balanced, atomic set of postings, recorded exactly once. Double-entry accounting is the baseline: every debit has a matching credit, and the transaction either commits in full or not at all.
Most homegrown ledgers miss the "exactly once" guarantee. Webhooks get redelivered, settlement files get re-uploaded, and workers retry after a crash between the rail confirmation and the ledger commit. Deduplication prevents the same event from posting twice.
The ledger enforces exactly-once with idempotency keys derived from the rail event's own identity. Retrying the same event results in the same posting instead of a new one. Without idempotency keys, reconciliation produces duplicate postings.
Any accounting period can be snapshotted as-of any timestamp
Any accounting period can be re-derived as-of any timestamp because the ledger stores two dates on every posting. When the event happened (effective date) and when the ledger recorded it (recorded date). The bi-temporality model keeps both as first-class fields.
A PSP settlement for August arrives on September 3, three days after the August 31 cutoff. It posts with an effective date in August and a recorded date in September.
The transaction is expressed in Numscript, Formance's purpose-built language for money that both engineering and finance teams can read and validate:
Query August as-of the September 3 recorded date, and the settlement is there. Query August as the ledger knew it on September 1, and it is not.
The ERP downstream picks up the adjustment dated in the open period, so nobody has to reopen the closed batch.
Corrections use new postings, never edits to existing rows
A correction is a new pair of postings that reverses the original and re-posts the right amount. The original stays where it is. If the August fee turns out to be $72,000.00 instead of $69,600.00, add the correction with its own effective and recorded timestamps.
The immutability guarantee means the August snapshot produced on September 3 still looks the same next March. Every transaction is hash-chained to the previous one, so any later tampering shows up when the chain is verified against a stored digest.
External balances reconcile continuously with owned exceptions
External balances only stay trustworthy if the ledger checks them against the outside world on a schedule the finance team can rely on.
The ledger compares its own balances against the balances reported by the providers holding the funds, running as often as the rail and reconciliation rules require.
When the two diverge, an alert fires and the exception enters the queue with a severity, owner, age, and close-blocking flag. By cutoff, the finance team is working through a short list of owned items instead of opening an investigation.
The key benefits of continuous accounting
Continuous accounting delivers four benefits: the close becomes a query, errors are caught in minutes, every exception gets tagged for action, and any accounting period is reproducible on demand.
1. Turns the month-end close into a query
50% of finance teams take six or more business days to close, and cash reconciliation alone takes 20–50 hours each month.
Continuous accounting turns the month-end close into a query because postings balance on write and external balances reconcile continuously, so reconciliation work happens in the background instead of the last week of the month.
The queue of blocking items is short or empty by cutoff. The close is a snapshot query against the effective-date index.
2. Catches ledger errors within minutes
Continuous validation on every write catches ledger errors within minutes because a system-level check runs before anyone sees the entry.
A batch-checked ledger only catches the same error at the next reconciliation, days or weeks later, after the variance has already spread into downstream reports.
Move the check into the ledger, and errors become alerts for the engineering and finance teams.
3. Every exception gets tagged for action at detection
Tag every drift with a severity, owner, age, and close-blocking flag the moment it is detected, so no reconciliation exception sits unassigned long enough to become a loss.
Without those four attributes, an exception can sit in a queue for months while nobody knows who should act on it, how urgent it is, or whether it blocks the close.
Attach the tags at detection time, and the queue stops being a backlog and starts being a work list.
4. Any accounting period is reproducible on demand
Auditors can reproduce any accounting period as-of any date, months or years later, because the ledger includes when the event happened and when the ledger learned about it.
Filter by the first to see what economically belongs in a period, then filter by the second to see what the ledger knew at any given moment.
Both timestamps live as first-class fields so that you can reproduce any period without a batch job or a reopened journal.
Why continuous accounting needs a ledger
Exactly-once posting, as-of snapshots, correction-by-posting, and continuous reconciliation with owned exceptions only hold if they live in the ledger database itself.
There are four reasons: the ledger must be the system of record, application-layer enforcement breaks under contention, the ERP general ledger sits downstream, and batch checks otherwise rely on tired humans.
1. The ledger must be the system of record
The core ledger must be the system of record for postings and balances, with reconciliation establishing how ledger records compare with external sources.
When the finance team reconciles from exports, the export becomes the de facto record, and the ledger's correctness rests on a spreadsheet rather than on the ledger itself.
Continuous accounting must live in the ledger database that records fund flows as they happen. Any reporting tool or reconciliation script bolted on downstream inherits the batch ledger's discrepancies.
2. Application-layer enforcement breaks under contention
When two writes hit the same account at the same time, application-layer enforcement breaks first. A balance the ledger cannot guarantee is a balance someone has to work out by hand at close.
The fix is to enforce the rules in the ledger itself. Both posting rules and period-end control rules follow the same principle: rules the ledger enforces are rules nobody has to re-check at month-end.
3. The ERP general ledger is downstream and not the source of truth
Teams sometimes assume the ERP general ledger is where financial truth lives, but the ERP is a reporting destination.
The general ledger in your ERP (NetSuite, SAP, or similar) sits downstream of the core ledger. It receives aggregated journals from the core ledger and owns the P&L.
That order matters. Continuous accounting properties must hold upstream, in the core ledger, because the ERP journal is only as correct as the file it was given. If the upstream ledger drifts, the ERP drifts with it, and no amount of month-end journal cleanup fixes that.
4. Batch checks depend on tired humans
Under capacity pressure, a third of accountants make several errors a week, and each one lands in the same period-end queue as the variances they were hired to find.
Batch checks concentrate the entire month's worth of verification work into the last few days of the period. Reviewers open exports, chase variances across systems, and reconcile balances against provider statements, all under a fixed deadline.
The volume is manageable when everything matches, but the team must investigate, explain, and correct every drift before the books can close. The queue grows faster than the team can drain it, and fatigue-driven errors become next month's issues.
Does your team need continuous accounting?
Continuous accounting is worth adopting the moment a settlement arrives three days after cutoff, an engineer traces a five-figure gap through application code, or a close is held open and then reopened for late chargebacks.
Run through the five checks below to see where your current close stands. If you check three or more, moving to continuous accounting will help.
1Your close takes more than five business days, and you spend most of that time chasing variances rather than reviewing them.
2A settlement or webhook that arrives after cutoff forces someone to reopen a closed journal batch.
3You reconcile from exports (CSV, spreadsheets, and BI dashboards) instead of directly against the ledger.
4A five-figure discrepancy takes an engineer more than a day to trace because the fee or FX logic lives in application code and not in the ledger.
5You cannot reproduce what the ledger looked like on a specific date last quarter without running a batch job or reopening a period.
Start with the ledger itself. Every check is either enforced by the ledger or worked around by humans, and moving enforcement into the ledger turns the close from a project into a query.