Reconciliation workflow

Two statements. One amber line. Close faster.

Reconcilia Cloud runs the daily match for Singapore finance teams between their bank feed, their ledger and their GST position — and leaves the approval where it belongs, with a person who signs it.

Tour Seven steps from feed to signed-off close
  1. Bank feed Connect and import
  2. Normalise Clean the rows
  3. Matching Accept the confident band
  4. Review Clear exceptions
Bank feed side Ledger side
  • DBS current account — feed row 1,840.20
  • 02 UOB collection — card settlement 312.75
  • 03 OCBC — supplier payment reference 4,096.00
  • Sales invoice — matched to feed row 1,840.20
  • 02 Card clearing account 312.75
  • 03 Trade creditor ledger line 4,096.00

Amount, date proximity and reference agree — proposed for approval, not posted.

Where automation ends

Three things the platform does not decide for you

  • Match proposals with visible evidence Automated
  • Approval of every posted match Named person
  • Accounting treatment of genuine differences Your team

Matching is the easy part to describe and the hard part to trust. The workflow is built so that speed never quietly becomes permission — every proposal carries the amounts, the date window and the reference it matched on, and nothing reaches the ledger until someone accepts it.

Feeds in from DBS UOB OCBC IRAS Full feed coverage

The sequence

From bank feed to signed-off close

This is the order a Singapore finance team actually works in. Seven steps, each one telling you what the platform handles and what still waits for a person.

Step one

Connect and import

You authorise a bank feed for each account you want reconciled. For the first import you choose a start date, and most teams pick the beginning of the previous month so the platform's output can be checked against a close they already finished manually.

Supplementary accounts come in through statement import. A separate account is not a reason to leave part of the reconciliation outside the workspace.

A hand lifting a printed bank statement from a laminate desk while a phone lies face down beside it
First import — statement pulled off the desk

Step two

Normalise and categorise

Incoming rows are cleaned before matching: dates parsed, reference strings trimmed of bank formatting noise, counterparty names extracted where the feed provides them. Currency and account context are attached to every row.

This step is invisible in the interface, which is the point. Normalisation is what makes the matching score meaningful instead of a string comparison that fails on an extra space.

  • Dates Parsed from the bank's own date format into one comparable value.
  • References Trimmed of formatting noise so the same reference matches twice.
  • Counterparties Extracted where the feed provides a name, ignored where it does not.
  • Context Currency and account attached to every row before it is scored.
Two ruled columns on a matte ledger page converging at a single centre rule with one small amber reflection at the join
Two sides of one entry, locked at a single rule

Step three

Automated matching

The platform proposes pairs. Amount equality or near-equality carries the most weight, then date proximity inside a configurable window, then reference and counterparty signals. Each proposal arrives with its evidence visible.

You can accept in bulk for the confident band and review the rest individually. Splitting that decision is what keeps the process fast without turning it into rubber-stamping.

Signals, in the order they are weighed
  1. Amount equality or near-equalityHeaviest
  2. Date proximity inside a configurable windowNext
  3. Reference and counterparty signalsSupporting

Step four

Clearing exceptions

What remains is grouped by cause. Each cleared exception teaches the platform a pattern, so the same supplier reference next month is more likely to match on its own. That is the compounding effect worth watching in your first quarter.

Cause Split payments

One bank line, several ledger entries. Matched against each entry in turn.

Cause Bank charges

Coded against an existing rule, or flagged when the rule no longer fits.

Cause FX differences

Assigned to gain or loss, with the rate context visible on the row.

Cause Duplicates

Flagged for removal upstream rather than reconciled twice on the ledger.

The same motion, six ways

Where a person still signs their name

Six moments of the same sequence — statement off the desk, the lock at the rule, the exception folder, the lines entered by hand, the close, and the review that keeps the audit trail honest.

A hand lifting a bank statement off a laminate desk with a phone turned face down beside it
Feed lands — statement off the desk
Two ruled columns on a matte ledger page converging at one rule with a small amber dot at the join
The match — two columns lock
An amber folder laid on a white laminate desk with one corner lifting in daylight
Exception flagged for review
Two hands entering figures into a ruled ledger book on a laminate desk with a calculator out of focus
Lines entered by hand at review
A closed ruled ledger book on a laminate desk with an amber pen laid diagonally across it
Close — ledger shut for the month
Three paper-clipped bundles of banknotes laid in ruled rows on a laminate desk with faces turned away
Multi-currency lines, one ledger

Six frames, one workflow — you can run all of them in your first month.

After the match

Adjustments, a parallel run, and the cadence that follows

Matching is the loud part of the workflow. Three quieter practices decide whether your first quarter of it goes well — how differences become journals, how the first month runs beside your old process, and what the calendar looks like after that.

Step five

Adjustments

Genuine differences produce a correcting journal with a reason code attached. Reason codes are chosen from your own list, which means the monthly report reflects your categories rather than a generic taxonomy.

Reviewing reason codes monthly is the cheapest process improvement available to a small finance team: it points at the source of most recurring exceptions.

Your own reasons, not ours

First month

Parallel run

Run the platform alongside your existing process for one cycle. Compare matched volume, exception categories, and the time the close took.

Most teams find the useful output of a parallel run is not the time saved but the exception report. It tells you which upstream habit was costing you the most hours.

Compare — matched volume · exception mix · close time

From month two

Recurring cadence

Daily feed lands overnight. Weekly, someone clears the exception queue, which keeps it to a manageable size. Monthly, GST review, close, and lock.

Two people can run this cadence for a small SME. The work is scheduled instead of accumulated, which is the difference between a close that finishes on day three and one that runs into the following week.

Daily · Weekly · Monthly · Lock
A matte white powder-coated steel panel with one warm amber reflection at the left edge
The surface a close leaves behind

Steps six and seven

GST review, then a close you can lock

Step six

GST review

With the ledger matched, GST treatment is reviewed at line level. Entries carrying standard-rated tax, zero-rated exports, and out-of-scope items are distinguishable in the working view.

This review is quick when the coding is clean and revealing when it is not. Either way, it happens before the return is built rather than after.

Review GST and IRAS support

Step seven

Month-end close

The close is a state rather than a document. When every account shows reconciled, the period is lockable, and the GST position for the period is final for reporting purposes.

Locking a period protects it. Later edits require an explicit unlock with a reason, which keeps your audit trail meaningful.

On purpose

What stays manual, on purpose

Someone approves matches. That is not a limitation to work around; it is the control that makes the automation auditable. A tool that posts matches without a named approver creates a problem the first time an auditor asks.

Splits, genuine FX differences, and miscoded entries also stay with a person. Judgement about accounting treatment belongs to your team, not to a scoring model.

The tool proposes. A person signs. That sentence is the whole control, and it is the one an auditor will ask you to prove.

Subscription

One monthly seat, the whole seven-step workflow

A monthly subscription for the finance team that runs reconciliation in-house. Pick the payment frequency that matches your own calendar.

Monthly seat Rolling

Monthly subscription

Billed each month per workspace

  • Connect and import bank feedsIncluded
  • Automated matching proposalsIncluded
  • Exception queue and reason codesIncluded
  • GST working view and close lockIncluded
  • Parallel-run comparison supportIncluded
See the plan details
Multi-account For teams of three to ten

Add seats as you go

Each seat runs the same workflow

  • Named approver per posted matchIncluded
  • Multi-currency ledger matchingIncluded
  • Monthly close and period lockIncluded
  • Reason-code reporting by categoryIncluded
  • Support during the parallel runIncluded
Compare plan options

Prices and included limits are listed on the pricing page.

Privacy notice

We keep a few cookies so your bank feed stays signed in and your reports render right. What we store